diff options
Diffstat (limited to 'apt-pkg/contrib')
34 files changed, 12603 insertions, 0 deletions
diff --git a/apt-pkg/contrib/arfile.cc b/apt-pkg/contrib/arfile.cc new file mode 100644 index 0000000..6d4a1f1 --- /dev/null +++ b/apt-pkg/contrib/arfile.cc @@ -0,0 +1,179 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + AR File - Handle an 'AR' archive + + AR Archives have plain text headers at the start of each file + section. The headers are aligned on a 2 byte boundary. + + Information about the structure of AR files can be found in ar(5) + on a BSD system, or in the binutils source. + + ##################################################################### */ + /*}}}*/ +// Include Files /*{{{*/ +#include <config.h> + +#include <apt-pkg/arfile.h> +#include <apt-pkg/error.h> +#include <apt-pkg/fileutl.h> +#include <apt-pkg/strutl.h> + +#include <string> +#include <string.h> +#include <sys/types.h> + +#include <apti18n.h> + /*}}}*/ + +struct ARArchive::MemberHeader +{ + char Name[16]; + char MTime[12]; + char UID[6]; + char GID[6]; + char Mode[8]; + char Size[10]; + char Magic[2]; +}; + +// ARArchive::ARArchive - Constructor /*{{{*/ +// --------------------------------------------------------------------- +/* */ +ARArchive::ARArchive(FileFd &File) : List(0), File(File) +{ + LoadHeaders(); +} + /*}}}*/ +// ARArchive::~ARArchive - Destructor /*{{{*/ +// --------------------------------------------------------------------- +/* */ +ARArchive::~ARArchive() +{ + while (List != 0) + { + Member *Tmp = List; + List = List->Next; + delete Tmp; + } +} + /*}}}*/ +// ARArchive::LoadHeaders - Load the headers from each file /*{{{*/ +// --------------------------------------------------------------------- +/* AR files are structured with a 8 byte magic string followed by a 60 + byte plain text header then the file data, another header, data, etc */ +bool ARArchive::LoadHeaders() +{ + off_t Left = File.Size(); + + // Check the magic byte + char Magic[8]; + if (File.Read(Magic,sizeof(Magic)) == false) + return false; + if (memcmp(Magic,"!<arch>\012",sizeof(Magic)) != 0) + return _error->Error(_("Invalid archive signature")); + Left -= sizeof(Magic); + + // Read the member list + while (Left > 0) + { + MemberHeader Head; + if (File.Read(&Head,sizeof(Head)) == false) + return _error->Error(_("Error reading archive member header")); + Left -= sizeof(Head); + + // Convert all of the integer members + Member *Memb = new Member(); + if (StrToNum(Head.MTime,Memb->MTime,sizeof(Head.MTime)) == false || + StrToNum(Head.UID,Memb->UID,sizeof(Head.UID)) == false || + StrToNum(Head.GID,Memb->GID,sizeof(Head.GID)) == false || + StrToNum(Head.Mode,Memb->Mode,sizeof(Head.Mode),8) == false || + StrToNum(Head.Size,Memb->Size,sizeof(Head.Size)) == false) + { + delete Memb; + return _error->Error(_("Invalid archive member header")); + } + + if (Left < 0 || Memb->Size > static_cast<unsigned long long>(Left)) + { + delete Memb; + return _error->Error(_("Invalid archive member header")); + } + // Check for an extra long name string + if (memcmp(Head.Name,"#1/",3) == 0) + { + char S[300]; + unsigned long Len; + if (StrToNum(Head.Name+3,Len,sizeof(Head.Size)-3) == false || + Len >= sizeof(S)) + { + delete Memb; + return _error->Error(_("Invalid archive member header")); + } + + if (Len > Memb->Size) + { + delete Memb; + return _error->Error(_("Invalid archive member header")); + } + + if (File.Read(S,Len) == false) + { + delete Memb; + return false; + } + S[Len] = 0; + Memb->Name = S; + Memb->Size -= Len; + Left -= Len; + } + else + { + unsigned int I = sizeof(Head.Name) - 1; + for (; Head.Name[I] == ' ' || Head.Name[I] == '/'; I--) + { + if (I == 0) + { + delete Memb; + return _error->Error(_("Invalid archive member header")); + } + } + Memb->Name = std::string(Head.Name,I+1); + } + + // Account for the AR header alignment + off_t Skip = Memb->Size % 2; + + // Add it to the list + Memb->Next = List; + List = Memb; + Memb->Start = File.Tell(); + if (File.Skip(Memb->Size + Skip) == false) + return false; + if (Left < (off_t)(Memb->Size + Skip)) + return _error->Error(_("Archive is too short")); + Left -= Memb->Size + Skip; + } + if (Left != 0) + return _error->Error(_("Failed to read the archive headers")); + + return true; +} + /*}}}*/ +// ARArchive::FindMember - Find a name in the member list /*{{{*/ +// --------------------------------------------------------------------- +/* Find a member with the given name */ +const ARArchive::Member *ARArchive::FindMember(const char *Name) const +{ + const Member *Res = List; + while (Res != 0) + { + if (Res->Name == Name) + return Res; + Res = Res->Next; + } + + return 0; +} + /*}}}*/ diff --git a/apt-pkg/contrib/arfile.h b/apt-pkg/contrib/arfile.h new file mode 100644 index 0000000..9b13ed9 --- /dev/null +++ b/apt-pkg/contrib/arfile.h @@ -0,0 +1,66 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + AR File - Handle an 'AR' archive + + This is a reader for the usual 4.4 BSD AR format. It allows raw + stream access to a single member at a time. Basically all this class + provides is header parsing and verification. It is up to the client + to correctly make use of the stream start/stop points. + + ##################################################################### */ + /*}}}*/ +#ifndef PKGLIB_ARFILE_H +#define PKGLIB_ARFILE_H + +#include <apt-pkg/macros.h> +#include <string> + +class FileFd; + +class APT_PUBLIC ARArchive +{ + struct MemberHeader; + public: + struct Member; + + protected: + + // Linked list of members + Member *List; + + bool LoadHeaders(); + + public: + + // The stream file + FileFd &File; + + // Locate a member by name + const Member *FindMember(const char *Name) const; + inline Member *Members() { return List; } + + APT_PUBLIC explicit ARArchive(FileFd &File); + APT_PUBLIC ~ARArchive(); +}; + +// A member of the archive +struct ARArchive::Member +{ + // Fields from the header + std::string Name; + unsigned long MTime; + unsigned long UID; + unsigned long GID; + unsigned long Mode; + unsigned long long Size; + + // Location of the data. + unsigned long long Start; + Member *Next; + + Member() : Start(0), Next(0) {}; +}; + +#endif diff --git a/apt-pkg/contrib/cdromutl.cc b/apt-pkg/contrib/cdromutl.cc new file mode 100644 index 0000000..f402017 --- /dev/null +++ b/apt-pkg/contrib/cdromutl.cc @@ -0,0 +1,293 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + CDROM Utilities - Some functions to manipulate CDROM mounts. + + These are here for the cdrom method and apt-cdrom. + + ##################################################################### */ + /*}}}*/ +// Include Files /*{{{*/ +#include <config.h> + +#include <apt-pkg/cdromutl.h> +#include <apt-pkg/configuration.h> +#include <apt-pkg/error.h> +#include <apt-pkg/fileutl.h> +#include <apt-pkg/hashes.h> +#include <apt-pkg/strutl.h> + +#include <iostream> +#include <string> +#include <vector> +#include <dirent.h> +#include <fcntl.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <sys/stat.h> +#include <sys/statvfs.h> +#include <unistd.h> + +#include <apti18n.h> + /*}}}*/ + +using std::string; + +// IsMounted - Returns true if the mount point is mounted /*{{{*/ +// --------------------------------------------------------------------- +/* This is a simple algorithm that should always work, we stat the mount point + and the '..' file in the mount point and see if they are on the same device. + By definition if they are the same then it is not mounted. This should + account for symlinked mount points as well. */ +bool IsMounted(string &Path) +{ + if (Path.empty() == true) + return false; + + // Need that trailing slash for directories + if (Path[Path.length() - 1] != '/') + Path += '/'; + + // if the path has a ".disk" directory we treat it as mounted + // this way even extracted copies of disks are recognized + if (DirectoryExists(Path + ".disk/") == true) + return true; + + /* First we check if the path is actually mounted, we do this by + stating the path and the previous directory (careful of links!) + and comparing their device fields. */ + struct stat Buf,Buf2; + if (stat(Path.c_str(),&Buf) != 0 || + stat((Path + "../").c_str(),&Buf2) != 0) + return _error->Errno("stat",_("Unable to stat the mount point %s"),Path.c_str()); + + if (Buf.st_dev == Buf2.st_dev) + return false; + return true; +} + /*}}}*/ +// UnmountCdrom - Unmount a cdrom /*{{{*/ +// --------------------------------------------------------------------- +/* Forking umount works much better than the umount syscall which can + leave /etc/mtab inconsistent. We drop all messages this produces. */ +bool UnmountCdrom(string Path) +{ + // do not generate errors, even if the mountpoint does not exist + // the mountpoint might be auto-created by the mount command + // and a non-existing mountpoint is surely not mounted + _error->PushToStack(); + bool const mounted = IsMounted(Path); + _error->RevertToStack(); + if (mounted == false) + return true; + + for (int i=0;i<3;i++) + { + + int Child = ExecFork(); + + // The child + if (Child == 0) + { + // Make all the fds /dev/null + int const null_fd = open("/dev/null",O_RDWR); + for (int I = 0; I != 3; ++I) + dup2(null_fd, I); + + if (_config->Exists("Acquire::cdrom::"+Path+"::UMount") == true) + { + if (system(_config->Find("Acquire::cdrom::"+Path+"::UMount").c_str()) != 0) + _exit(100); + _exit(0); + } + else + { + const char * const Args[] = { + "umount", + Path.c_str(), + nullptr + }; + execvp(Args[0], const_cast<char **>(Args)); + _exit(100); + } + } + + // if it can not be umounted, give it a bit more time + // this can happen when auto-mount magic or fs/cdrom prober attack + if (ExecWait(Child,"umount",true) == true) + return true; + sleep(1); + } + + return false; +} + /*}}}*/ +// MountCdrom - Mount a cdrom /*{{{*/ +// --------------------------------------------------------------------- +/* We fork mount and drop all messages */ +bool MountCdrom(string Path, string DeviceName) +{ + // do not generate errors, even if the mountpoint does not exist + // the mountpoint might be auto-created by the mount command + _error->PushToStack(); + bool const mounted = IsMounted(Path); + _error->RevertToStack(); + if (mounted == true) + return true; + + int Child = ExecFork(); + + // The child + if (Child == 0) + { + // Make all the fds /dev/null + int const null_fd = open("/dev/null",O_RDWR); + for (int I = 0; I != 3; ++I) + dup2(null_fd, I); + + if (_config->Exists("Acquire::cdrom::"+Path+"::Mount") == true) + { + if (system(_config->Find("Acquire::cdrom::"+Path+"::Mount").c_str()) != 0) + _exit(100); + _exit(0); + } + else + { + const char *Args[10]; + Args[0] = "mount"; + if (DeviceName == "") + { + Args[1] = Path.c_str(); + Args[2] = 0; + } else { + Args[1] = DeviceName.c_str(); + Args[2] = Path.c_str(); + Args[3] = 0; + } + execvp(Args[0],(char **)Args); + _exit(100); + } + } + + // Wait for mount + return ExecWait(Child,"mount",true); +} + /*}}}*/ +// IdentCdrom - Generate a unique string for this CD /*{{{*/ +// --------------------------------------------------------------------- +/* We convert everything we hash into a string, this prevents byte size/order + from effecting the outcome. */ +bool IdentCdrom(string CD,string &Res,unsigned int Version) +{ + Hashes Hash(Hashes::MD5SUM); + bool writable_media = false; + + int dirfd = open(CD.c_str(), O_RDONLY | O_DIRECTORY | O_CLOEXEC); + if (dirfd == -1) + return _error->Errno("open",_("Unable to read %s"),CD.c_str()); + + // if we are on a writable medium (like a usb-stick) that is just + // used like a cdrom don't use "." as it will constantly change, + // use .disk instead + if (faccessat(dirfd, ".", W_OK, 0) == 0) + { + int diskfd = openat(dirfd, "./.disk", O_RDONLY | O_DIRECTORY | O_CLOEXEC, 0); + if (diskfd != -1) + { + close(dirfd); + dirfd = diskfd; + writable_media = true; + CD = CD.append("/.disk"); + if (_config->FindB("Debug::aptcdrom",false) == true) + std::clog << "Found writable cdrom, using alternative path: " << CD + << std::endl; + } + } + + DIR * const D = fdopendir(dirfd); + if (D == nullptr) + return _error->Errno("opendir",_("Unable to read %s"),CD.c_str()); + + /* Run over the directory, we assume that the reader order will never + change as the media is read-only. In theory if the kernel did + some sort of wacked caching this might not be true.. */ + for (struct dirent *Dir = readdir(D); Dir != 0; Dir = readdir(D)) + { + // Skip some files.. + if (strcmp(Dir->d_name,".") == 0 || + strcmp(Dir->d_name,"..") == 0) + continue; + + std::string S; + if (Version <= 1) + S = std::to_string(Dir->d_ino); + else + { + struct stat Buf; + if (fstatat(dirfd, Dir->d_name, &Buf, 0) != 0) + continue; + S = std::to_string(Buf.st_mtime); + } + + Hash.Add(S.c_str()); + Hash.Add(Dir->d_name); + } + + // Some stats from the fsys + std::string S; + if (_config->FindB("Debug::identcdrom",false) == false) + { + struct statvfs Buf; + if (fstatvfs(dirfd, &Buf) != 0) + return _error->Errno("statfs",_("Failed to stat the cdrom")); + + // We use a kilobyte block size to avoid overflow + S = std::to_string(Buf.f_blocks * (Buf.f_bsize / 1024)); + if (writable_media == false) + S.append(" ").append(std::to_string(Buf.f_bfree * (Buf.f_bsize / 1024))); + Hash.Add(S.c_str(), S.length()); + strprintf(S, "-%u", Version); + } + else + strprintf(S, "-%u.debug", Version); + + closedir(D); + Res = Hash.GetHashString(Hashes::MD5SUM).HashValue().append(std::move(S)); + return true; +} + /*}}}*/ +// FindMountPointForDevice - Find mountpoint for the given device /*{{{*/ +string FindMountPointForDevice(const char *devnode) +{ + // this is the order that mount uses as well + std::vector<std::string> const mounts = _config->FindVector("Dir::state::MountPoints", "/etc/mtab,/proc/mount"); + + for (std::vector<std::string>::const_iterator m = mounts.begin(); m != mounts.end(); ++m) + if (FileExists(*m) == true) + { + char * line = NULL; + size_t line_len = 0; + FILE * f = fopen(m->c_str(), "r"); + while(getline(&line, &line_len, f) != -1) + { + char * out[] = { NULL, NULL, NULL }; + TokSplitString(' ', line, out, 3); + if (out[2] != NULL || out[1] == NULL || out[0] == NULL) + continue; + if (strcmp(out[0], devnode) != 0) + continue; + fclose(f); + // unescape the \0XXX chars in the path + string mount_point = out[1]; + free(line); + return DeEscapeString(mount_point); + } + fclose(f); + free(line); + } + + return string(); +} + /*}}}*/ diff --git a/apt-pkg/contrib/cdromutl.h b/apt-pkg/contrib/cdromutl.h new file mode 100644 index 0000000..1384cea --- /dev/null +++ b/apt-pkg/contrib/cdromutl.h @@ -0,0 +1,24 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + CDROM Utilities - Some functions to manipulate CDROM mounts. + + ##################################################################### */ + /*}}}*/ +#ifndef PKGLIB_CDROMUTL_H +#define PKGLIB_CDROMUTL_H + +#include <apt-pkg/macros.h> + +#include <string> + + +// mount cdrom, DeviceName (e.g. /dev/sr0) is optional +APT_PUBLIC bool MountCdrom(std::string Path, std::string DeviceName=""); +APT_PUBLIC bool UnmountCdrom(std::string Path); +APT_PUBLIC bool IdentCdrom(std::string CD,std::string &Res,unsigned int Version = 2); +APT_PUBLIC bool IsMounted(std::string &Path); +APT_PUBLIC std::string FindMountPointForDevice(const char *device); + +#endif diff --git a/apt-pkg/contrib/cmndline.cc b/apt-pkg/contrib/cmndline.cc new file mode 100644 index 0000000..998af2f --- /dev/null +++ b/apt-pkg/contrib/cmndline.cc @@ -0,0 +1,445 @@ +// -*- mode: cpp; mode: fold -*- +// SPDX-License-Identifier: GPL-2.0+ +// Description /*{{{*/ +/* ###################################################################### + + Command Line Class - Sophisticated command line parser + + This file had this historic note, but now includes further changes + under the GPL-2.0+: + + This source is placed in the Public Domain, do with it what you will + It was originally written by Jason Gunthorpe <jgg@debian.org>. + + ##################################################################### */ + /*}}}*/ +// Include files /*{{{*/ +#include <config.h> + +#include <apt-pkg/cmndline.h> +#include <apt-pkg/configuration.h> +#include <apt-pkg/error.h> +#include <apt-pkg/strutl.h> + +#include <string> +#include <stddef.h> +#include <stdlib.h> +#include <string.h> + +#include <apti18n.h> + /*}}}*/ +using namespace std; + +// CommandLine::CommandLine - Constructor /*{{{*/ +// --------------------------------------------------------------------- +/* */ +CommandLine::CommandLine(Args *AList,Configuration *Conf) : ArgList(AList), + Conf(Conf), FileList(0) +{ +} +CommandLine::CommandLine() : ArgList(NULL), Conf(NULL), FileList(0) +{ +} + /*}}}*/ +// CommandLine::~CommandLine - Destructor /*{{{*/ +// --------------------------------------------------------------------- +/* */ +CommandLine::~CommandLine() +{ + delete [] FileList; +} + /*}}}*/ +// CommandLine::GetCommand - return the first non-option word /*{{{*/ +char const * CommandLine::GetCommand(Dispatch const * const Map, + unsigned int const argc, char const * const * const argv) +{ + // if there is a -- on the line there must be the word we search for either + // before it (as -- marks the end of the options) or right after it (as we can't + // decide if the command is actually an option, given that in theory, you could + // have parameters named like commands) + for (size_t i = 1; i < argc; ++i) + { + if (strcmp(argv[i], "--") != 0) + continue; + // check if command is before -- + for (size_t k = 1; k < i; ++k) + for (size_t j = 0; Map[j].Match != NULL; ++j) + if (strcmp(argv[k], Map[j].Match) == 0) + return Map[j].Match; + // see if the next token after -- is the command + ++i; + if (i < argc) + for (size_t j = 0; Map[j].Match != NULL; ++j) + if (strcmp(argv[i], Map[j].Match) == 0) + return Map[j].Match; + // we found a --, but not a command + return NULL; + } + // no --, so search for the first word matching a command + // FIXME: How like is it that an option parameter will be also a valid Match ? + for (size_t i = 1; i < argc; ++i) + { + if (*(argv[i]) == '-') + continue; + for (size_t j = 0; Map[j].Match != NULL; ++j) + if (strcmp(argv[i], Map[j].Match) == 0) + return Map[j].Match; + } + return NULL; +} + /*}}}*/ +// CommandLine::Parse - Main action member /*{{{*/ +// --------------------------------------------------------------------- +/* */ +bool CommandLine::Parse(int argc,const char **argv) +{ + delete [] FileList; + FileList = new const char *[argc]; + const char **Files = FileList; + int I; + for (I = 1; I != argc; I++) + { + const char *Opt = argv[I]; + + // It is not an option + if (*Opt != '-') + { + *Files++ = Opt; + continue; + } + + Opt++; + + // Double dash signifies the end of option processing + if (*Opt == '-' && Opt[1] == 0) + { + I++; + break; + } + + // Single dash is a short option + if (*Opt != '-') + { + // Iterate over each letter + while (*Opt != 0) + { + // Search for the option + Args *A; + for (A = ArgList; A->end() == false && A->ShortOpt != *Opt; A++); + if (A->end() == true) + return _error->Error(_("Command line option '%c' [from %s] is not understood in combination with the other options."),*Opt,argv[I]); + + if (HandleOpt(I,argc,argv,Opt,A) == false) + return false; + if (*Opt != 0) + Opt++; + } + continue; + } + + Opt++; + + // Match up to a = against the list + Args *A; + const char *OptEnd = strchrnul(Opt, '='); + for (A = ArgList; A->end() == false && + (A->LongOpt == 0 || stringcasecmp(Opt,OptEnd,A->LongOpt) != 0); + ++A); + + // Failed, look for a word after the first - (no-foo) + bool PreceedMatch = false; + if (A->end() == true) + { + Opt = (const char*) memchr(Opt, '-', OptEnd - Opt); + if (Opt == NULL) + return _error->Error(_("Command line option %s is not understood in combination with the other options"),argv[I]); + Opt++; + + for (A = ArgList; A->end() == false && + (A->LongOpt == 0 || stringcasecmp(Opt,OptEnd,A->LongOpt) != 0); + ++A); + + // Failed again.. + if (A->end() == true && OptEnd - Opt != 1) + return _error->Error(_("Command line option %s is not understood in combination with the other options"),argv[I]); + + // The option could be a single letter option prefixed by a no-.. + if (A->end() == true) + { + for (A = ArgList; A->end() == false && A->ShortOpt != *Opt; A++); + + if (A->end() == true) + return _error->Error(_("Command line option %s is not understood in combination with the other options"),argv[I]); + } + + // The option is not boolean + if (A->IsBoolean() == false) + return _error->Error(_("Command line option %s is not boolean"),argv[I]); + PreceedMatch = true; + } + + // Deal with it. + OptEnd--; + if (HandleOpt(I,argc,argv,OptEnd,A,PreceedMatch) == false) + return false; + } + + // Copy any remaining file names over + for (; I != argc; I++) + *Files++ = argv[I]; + *Files = 0; + + SaveInConfig(argc, argv); + + return true; +} + /*}}}*/ +// CommandLine::HandleOpt - Handle a single option including all flags /*{{{*/ +// --------------------------------------------------------------------- +/* This is a helper function for parser, it looks at a given argument + and looks for specific patterns in the string, it gets tokanized + -ruffly- like -*[yes|true|enable]-(o|longopt)[=][ ][argument] */ +bool CommandLine::HandleOpt(int &I,int argc,const char *argv[], + const char *&Opt,Args *A,bool PreceedMatch) +{ + const char *Argument = 0; + bool CertainArg = false; + int IncI = 0; + + /* Determine the possible location of an option or 0 if their is + no option */ + if (Opt[1] == 0) + { + if (I + 1 < argc && argv[I+1][0] != '-') + Argument = argv[I+1]; + + IncI = 1; + } + else + { + if (Opt[1] == '=') + { + CertainArg = true; + Argument = Opt + 2; + } + else + Argument = Opt + 1; + } + + // Option is an argument set + if ((A->Flags & HasArg) == HasArg) + { + if (Argument == 0) + return _error->Error(_("Option %s requires an argument."),argv[I]); + Opt += strlen(Opt); + I += IncI; + + // Parse a configuration file + if ((A->Flags & ConfigFile) == ConfigFile) + return ReadConfigFile(*Conf,Argument); + + // Arbitrary item specification + if ((A->Flags & ArbItem) == ArbItem) + { + const char * const J = strchr(Argument, '='); + if (J == nullptr) + return _error->Error(_("Option %s: Configuration item specification must have an =<val>."),argv[I]); + + Conf->Set(string(Argument,J-Argument), J+1); + return true; + } + + const char *I = strchrnul(A->ConfName, ' '); + if (*I == ' ') + Conf->Set(string(A->ConfName,0,I-A->ConfName),string(I+1) + Argument); + else + Conf->Set(A->ConfName,string(I) + Argument); + + return true; + } + + // Option is an integer level + if ((A->Flags & IntLevel) == IntLevel) + { + // There might be an argument + if (Argument != 0) + { + char *EndPtr; + unsigned long Value = strtol(Argument,&EndPtr,10); + + // Conversion failed and the argument was specified with an =s + if (EndPtr == Argument && CertainArg == true) + return _error->Error(_("Option %s requires an integer argument, not '%s'"),argv[I],Argument); + + // Conversion was ok, set the value and return + if (EndPtr != 0 && EndPtr != Argument && *EndPtr == 0) + { + Conf->Set(A->ConfName,Value); + Opt += strlen(Opt); + I += IncI; + return true; + } + } + + // Increase the level + Conf->Set(A->ConfName,Conf->FindI(A->ConfName)+1); + return true; + } + + // Option is a boolean + int Sense = -1; // -1 is unspecified, 0 is yes 1 is no + + // Look for an argument. + while (1) + { + // Look at preceding text + char Buffer[300]; + if (Argument == 0) + { + if (PreceedMatch == false) + break; + + if (strlen(argv[I]) >= sizeof(Buffer)) + return _error->Error(_("Option '%s' is too long"),argv[I]); + + // Skip the leading dash + const char *J = argv[I]; + for (; *J == '-'; J++) + ; + + const char *JEnd = strchr(J, '-'); + if (JEnd != NULL) + { + strncpy(Buffer,J,JEnd - J); + Buffer[JEnd - J] = 0; + Argument = Buffer; + CertainArg = true; + } + else + break; + } + + // Check for boolean + Sense = StringToBool(Argument); + if (Sense >= 0) + { + // Eat the argument + if (Argument != Buffer) + { + Opt += strlen(Opt); + I += IncI; + } + break; + } + + if (CertainArg == true) + return _error->Error(_("Sense %s is not understood, try true or false."),Argument); + + Argument = 0; + } + + // Indeterminate sense depends on the flag + if (Sense == -1) + { + if ((A->Flags & InvBoolean) == InvBoolean) + Sense = 0; + else + Sense = 1; + } + + Conf->Set(A->ConfName,Sense); + return true; +} + /*}}}*/ +// CommandLine::FileSize - Count the number of filenames /*{{{*/ +// --------------------------------------------------------------------- +/* */ +unsigned int CommandLine::FileSize() const +{ + unsigned int Count = 0; + for (const char **I = FileList; I != 0 && *I != 0; I++) + Count++; + return Count; +} + /*}}}*/ +// CommandLine::DispatchArg - Do something with the first arg /*{{{*/ +bool CommandLine::DispatchArg(Dispatch const * const Map,bool NoMatch) +{ + int I; + for (I = 0; Map[I].Match != 0; I++) + { + if (strcmp(FileList[0],Map[I].Match) == 0) + { + bool Res = Map[I].Handler(*this); + if (Res == false && _error->PendingError() == false) + _error->Error("Handler silently failed"); + return Res; + } + } + + // No matching name + if (Map[I].Match == 0) + { + if (NoMatch == true) + _error->Error(_("Invalid operation %s"),FileList[0]); + } + + return false; +} + /*}}}*/ +// CommandLine::SaveInConfig - for output later in a logfile or so /*{{{*/ +// --------------------------------------------------------------------- +/* We save the commandline here to have it around later for e.g. logging. + It feels a bit like a hack here and isn't bulletproof, but it is better + than nothing after all. */ +void CommandLine::SaveInConfig(unsigned int const &argc, char const * const * const argv) +{ + char cmdline[100 + argc * 50]; + memset(cmdline, 0, sizeof(cmdline)); + unsigned int length = 0; + bool lastWasOption = false; + bool closeQuote = false; + for (unsigned int i = 0; i < argc && length < sizeof(cmdline); ++i, ++length) + { + for (unsigned int j = 0; argv[i][j] != '\0' && length < sizeof(cmdline)-2; ++j) + { + // we can't really sensibly deal with quoting so skip it + if (strchr("\"\'\r\n", argv[i][j]) != nullptr) + continue; + cmdline[length++] = argv[i][j]; + if (lastWasOption == true && argv[i][j] == '=') + { + // That is possibly an option: Quote it if it includes spaces, + // the benefit is that this will eliminate also most false positives + const char* c = strchr(&argv[i][j+1], ' '); + if (c == NULL) continue; + cmdline[length++] = '\''; + closeQuote = true; + } + } + if (closeQuote == true) + { + cmdline[length++] = '\''; + closeQuote = false; + } + // Problem: detects also --hello + if (cmdline[length-1] == 'o') + lastWasOption = true; + cmdline[length] = ' '; + } + cmdline[--length] = '\0'; + _config->Set("CommandLine::AsString", cmdline); +} + /*}}}*/ +CommandLine::Args CommandLine::MakeArgs(char ShortOpt, char const *LongOpt, char const *ConfName, unsigned long Flags)/*{{{*/ +{ + /* In theory, this should be a constructor for CommandLine::Args instead, + but this breaks compatibility as gcc thinks this is a c++11 initializer_list */ + CommandLine::Args arg; + arg.ShortOpt = ShortOpt; + arg.LongOpt = LongOpt; + arg.ConfName = ConfName; + arg.Flags = Flags; + return arg; +} + /*}}}*/ diff --git a/apt-pkg/contrib/cmndline.h b/apt-pkg/contrib/cmndline.h new file mode 100644 index 0000000..40d384f --- /dev/null +++ b/apt-pkg/contrib/cmndline.h @@ -0,0 +1,113 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + Command Line Class - Sophisticated command line parser + + This class provides a unified command line parser/option handliner/ + configuration mechanism. It allows the caller to specify the option + set and map the option set into the configuration class or other + special functioning. + + Filenames are stripped from the option stream and put into their + own array. + + The argument descriptor array can be initialized as: + + CommandLine::Args Args[] = + {{'q',"quiet","apt::get::quiet",CommandLine::IntLevel}, + {0,0,0,0}}; + + The flags mean, + HasArg - Means the argument has a value + IntLevel - Means the argument is an integer level indication, the + following -qqqq (+3) -q5 (=5) -q=5 (=5) are valid + Boolean - Means it is true/false or yes/no. + -d (true) --no-d (false) --yes-d (true) + --long (true) --no-long (false) --yes-long (true) + -d=yes (true) -d=no (false) Words like enable, disable, + true false, yes no and on off are recognized in logical + places. + InvBoolean - Same as boolean but the case with no specified sense + (first case) is set to false. + ConfigFile - Means this flag should be interprited as the name of + a config file to read in at this point in option processing. + Implies HasArg. + ArbItem - Means the item is an arbitrary configuration string of + the form item=value, where item is passed directly + to the configuration class. + The default, if the flags are 0 is to use Boolean + + ##################################################################### */ + /*}}}*/ +#ifndef PKGLIB_CMNDLINE_H +#define PKGLIB_CMNDLINE_H + +#include <apt-pkg/macros.h> + + +class Configuration; + +class APT_PUBLIC CommandLine +{ + public: + struct Args; + struct Dispatch; + struct DispatchWithHelp; + + protected: + + Args *ArgList; + Configuration *Conf; + bool HandleOpt(int &I,int argc,const char *argv[], + const char *&Opt,Args *A,bool PreceedeMatch = false); + void static SaveInConfig(unsigned int const &argc, char const * const * const argv); + + public: + + enum AFlags + { + HasArg = (1 << 0), + IntLevel = (1 << 1), + Boolean = (1 << 2), + InvBoolean = (1 << 3), + ConfigFile = (1 << 4) | HasArg, + ArbItem = (1 << 5) | HasArg + }; + + const char **FileList; + + bool Parse(int argc,const char **argv); + void ShowHelp(); + unsigned int FileSize() const APT_PURE; + bool DispatchArg(Dispatch const * const List,bool NoMatch = true); + + static char const * GetCommand(Dispatch const * const Map, + unsigned int const argc, char const * const * const argv) APT_PURE; + + static CommandLine::Args MakeArgs(char ShortOpt, char const *LongOpt, + char const *ConfName, unsigned long Flags) APT_PURE; + + CommandLine(); + CommandLine(Args *AList,Configuration *Conf); + ~CommandLine(); +}; + +struct CommandLine::Args +{ + char ShortOpt; + const char *LongOpt; + const char *ConfName; + unsigned long Flags; + + inline bool end() {return ShortOpt == 0 && LongOpt == 0;}; + inline bool IsBoolean() {return Flags == 0 || (Flags & (Boolean|InvBoolean)) != 0;}; +}; + +struct CommandLine::Dispatch +{ + const char *Match; + bool (*Handler)(CommandLine &); +}; + +#endif diff --git a/apt-pkg/contrib/configuration.cc b/apt-pkg/contrib/configuration.cc new file mode 100644 index 0000000..1134ed6 --- /dev/null +++ b/apt-pkg/contrib/configuration.cc @@ -0,0 +1,1194 @@ +// -*- mode: cpp; mode: fold -*- +// SPDX-License-Identifier: GPL-2.0+ +// Description /*{{{*/ +/* ###################################################################### + + Configuration Class + + This class provides a configuration file and command line parser + for a tree-oriented configuration environment. All runtime configuration + is stored in here. + + This file had this historic note, but now includes further changes + under the GPL-2.0+: + + This source is placed in the Public Domain, do with it what you will + It was originally written by Jason Gunthorpe <jgg@debian.org>. + + ##################################################################### */ + /*}}}*/ +// Include files /*{{{*/ +#include <config.h> + +#include <apt-pkg/configuration.h> +#include <apt-pkg/error.h> +#include <apt-pkg/fileutl.h> +#include <apt-pkg/macros.h> +#include <apt-pkg/strutl.h> +#include <apt-pkg/string_view.h> + +#include <ctype.h> +#include <regex.h> +#include <stddef.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +#include <algorithm> +#include <array> +#include <fstream> +#include <iterator> +#include <numeric> +#include <sstream> +#include <stack> +#include <string> +#include <unordered_map> +#include <vector> + +#include <apti18n.h> + +using namespace std; + /*}}}*/ + +Configuration *_config = new Configuration; + +/* TODO: This config verification shouldn't be using a static variable + but a Cnf-member – but that would need ABI breaks and stuff and for now + that really is an apt-dev-only tool, so it isn't that bad that it is + unusable and allaround a bit strange */ +enum class APT_HIDDEN ConfigType { UNDEFINED, INT, BOOL, STRING, STRING_OR_BOOL, STRING_OR_LIST, FILE, DIR, LIST, PROGRAM_PATH = FILE }; +APT_HIDDEN std::unordered_map<std::string, ConfigType> apt_known_config {}; +static std::string getConfigTypeString(ConfigType const type) /*{{{*/ +{ + switch (type) + { + case ConfigType::UNDEFINED: return "UNDEFINED"; + case ConfigType::INT: return "INT"; + case ConfigType::BOOL: return "BOOL"; + case ConfigType::STRING: return "STRING"; + case ConfigType::STRING_OR_BOOL: return "STRING_OR_BOOL"; + case ConfigType::FILE: return "FILE"; + case ConfigType::DIR: return "DIR"; + case ConfigType::LIST: return "LIST"; + case ConfigType::STRING_OR_LIST: return "STRING_OR_LIST"; + } + return "UNKNOWN"; +} + /*}}}*/ +static ConfigType getConfigType(std::string const &type) /*{{{*/ +{ + if (type == "<INT>") + return ConfigType::INT; + else if (type == "<BOOL>") + return ConfigType::BOOL; + else if (type == "<STRING>") + return ConfigType::STRING; + else if (type == "<STRING_OR_BOOL>") + return ConfigType::STRING_OR_BOOL; + else if (type == "<FILE>") + return ConfigType::FILE; + else if (type == "<DIR>") + return ConfigType::DIR; + else if (type == "<LIST>") + return ConfigType::LIST; + else if (type == "<STRING_OR_LIST>") + return ConfigType::STRING_OR_LIST; + else if (type == "<PROGRAM_PATH>") + return ConfigType::PROGRAM_PATH; + return ConfigType::UNDEFINED; +} + /*}}}*/ +// checkFindConfigOptionType - workhorse of option checking /*{{{*/ +static void checkFindConfigOptionTypeInternal(std::string name, ConfigType const type) +{ + std::transform(name.begin(), name.end(), name.begin(), ::tolower); + auto known = apt_known_config.find(name); + if (known == apt_known_config.cend()) + { + auto const rcolon = name.rfind(':'); + if (rcolon != std::string::npos) + { + known = apt_known_config.find(name.substr(0, rcolon) + ":*"); + if (known == apt_known_config.cend()) + { + auto const parts = StringSplit(name, "::"); + size_t psize = parts.size(); + if (psize > 1) + { + for (size_t max = psize; max != 1; --max) + { + std::ostringstream os; + std::copy(parts.begin(), parts.begin() + max, std::ostream_iterator<std::string>(os, "::")); + os << "**"; + known = apt_known_config.find(os.str()); + if (known != apt_known_config.cend() && known->second == ConfigType::UNDEFINED) + return; + } + for (size_t max = psize - 1; max != 1; --max) + { + std::ostringstream os; + std::copy(parts.begin(), parts.begin() + max - 1, std::ostream_iterator<std::string>(os, "::")); + os << "*::"; + std::copy(parts.begin() + max + 1, parts.end() - 1, std::ostream_iterator<std::string>(os, "::")); + os << *(parts.end() - 1); + known = apt_known_config.find(os.str()); + if (known != apt_known_config.cend()) + break; + } + } + } + } + } + if (known == apt_known_config.cend()) + _error->Warning("Using unknown config option »%s« of type %s", + name.c_str(), getConfigTypeString(type).c_str()); + else if (known->second != type) + { + if (known->second == ConfigType::DIR && type == ConfigType::FILE) + ; // implementation detail + else if (type == ConfigType::STRING && (known->second == ConfigType::FILE || known->second == ConfigType::DIR)) + ; // TODO: that might be an error or not, we will figure this out later + else if (known->second == ConfigType::STRING_OR_BOOL && (type == ConfigType::BOOL || type == ConfigType::STRING)) + ; + else if (known->second == ConfigType::STRING_OR_LIST && (type == ConfigType::LIST || type == ConfigType::STRING)) + ; + else + _error->Warning("Using config option »%s« of type %s as a type %s", + name.c_str(), getConfigTypeString(known->second).c_str(), getConfigTypeString(type).c_str()); + } +} +static void checkFindConfigOptionType(char const * const name, ConfigType const type) +{ + if (apt_known_config.empty()) + return; + checkFindConfigOptionTypeInternal(name, type); +} + /*}}}*/ +static bool LoadConfigurationIndex(std::string const &filename) /*{{{*/ +{ + apt_known_config.clear(); + if (filename.empty()) + return true; + Configuration Idx; + if (ReadConfigFile(Idx, filename) == false) + return false; + + Configuration::Item const * Top = Idx.Tree(nullptr); + if (unlikely(Top == nullptr)) + return false; + + do { + if (Top->Value.empty() == false) + { + std::string fulltag = Top->FullTag(); + std::transform(fulltag.begin(), fulltag.end(), fulltag.begin(), ::tolower); + apt_known_config.emplace(std::move(fulltag), getConfigType(Top->Value)); + } + + if (Top->Child != nullptr) + { + Top = Top->Child; + continue; + } + + while (Top != nullptr && Top->Next == nullptr) + Top = Top->Parent; + if (Top != nullptr) + Top = Top->Next; + } while (Top != nullptr); + + return true; +} + /*}}}*/ + +// Configuration::Configuration - Constructor /*{{{*/ +// --------------------------------------------------------------------- +/* */ +Configuration::Configuration() : ToFree(true) +{ + Root = new Item; +} +Configuration::Configuration(const Item *Root) : Root((Item *)Root), ToFree(false) +{ +} + /*}}}*/ +// Configuration::~Configuration - Destructor /*{{{*/ +// --------------------------------------------------------------------- +/* */ +Configuration::~Configuration() +{ + if (ToFree == false) + return; + + Item *Top = Root; + for (; Top != 0;) + { + if (Top->Child != 0) + { + Top = Top->Child; + continue; + } + + while (Top != 0 && Top->Next == 0) + { + Item *Parent = Top->Parent; + delete Top; + Top = Parent; + } + if (Top != 0) + { + Item *Next = Top->Next; + delete Top; + Top = Next; + } + } +} + /*}}}*/ +// Configuration::Lookup - Lookup a single item /*{{{*/ +// --------------------------------------------------------------------- +/* This will lookup a single item by name below another item. It is a + helper function for the main lookup function */ +Configuration::Item *Configuration::Lookup(Item *Head,const char *S, + unsigned long const &Len,bool const &Create) +{ + int Res = 1; + Item *I = Head->Child; + Item **Last = &Head->Child; + + // Empty strings match nothing. They are used for lists. + if (Len != 0) + { + for (; I != 0; Last = &I->Next, I = I->Next) + if (Len == I->Tag.length() && (Res = stringcasecmp(I->Tag,S,S + Len)) == 0) + break; + } + else + for (; I != 0; Last = &I->Next, I = I->Next); + + if (Res == 0) + return I; + if (Create == false) + return 0; + + I = new Item; + I->Tag.assign(S,Len); + I->Next = *Last; + I->Parent = Head; + *Last = I; + return I; +} + /*}}}*/ +// Configuration::Lookup - Lookup a fully scoped item /*{{{*/ +// --------------------------------------------------------------------- +/* This performs a fully scoped lookup of a given name, possibly creating + new items */ +Configuration::Item *Configuration::Lookup(const char *Name,bool const &Create) +{ + if (Name == 0) + return Root->Child; + + const char *Start = Name; + const char *End = Start + strlen(Name); + const char *TagEnd = Name; + Item *Itm = Root; + for (; End - TagEnd >= 2; TagEnd++) + { + if (TagEnd[0] == ':' && TagEnd[1] == ':') + { + Itm = Lookup(Itm,Start,TagEnd - Start,Create); + if (Itm == 0) + return 0; + TagEnd = Start = TagEnd + 2; + } + } + + // This must be a trailing ::, we create unique items in a list + if (End - Start == 0) + { + if (Create == false) + return 0; + } + + Itm = Lookup(Itm,Start,End - Start,Create); + return Itm; +} + /*}}}*/ +// Configuration::Find - Find a value /*{{{*/ +// --------------------------------------------------------------------- +/* */ +string Configuration::Find(const char *Name,const char *Default) const +{ + checkFindConfigOptionType(Name, ConfigType::STRING); + const Item *Itm = Lookup(Name); + if (Itm == 0 || Itm->Value.empty() == true) + { + if (Default == 0) + return ""; + else + return Default; + } + + return Itm->Value; +} + /*}}}*/ +// Configuration::FindFile - Find a Filename /*{{{*/ +// --------------------------------------------------------------------- +/* Directories are stored as the base dir in the Parent node and the + sub directory in sub nodes with the final node being the end filename + */ +string Configuration::FindFile(const char *Name,const char *Default) const +{ + checkFindConfigOptionType(Name, ConfigType::FILE); + const Item *RootItem = Lookup("RootDir"); + std::string result = (RootItem == 0) ? "" : RootItem->Value; + if(result.empty() == false && result[result.size() - 1] != '/') + result.push_back('/'); + + const Item *Itm = Lookup(Name); + if (Itm == 0 || Itm->Value.empty() == true) + { + if (Default != 0) + result.append(Default); + } + else + { + string val = Itm->Value; + while (Itm->Parent != 0) + { + if (Itm->Parent->Value.empty() == true) + { + Itm = Itm->Parent; + continue; + } + + // Absolute + if (val.length() >= 1 && val[0] == '/') + { + if (val.compare(0, 9, "/dev/null") == 0) + val.erase(9); + break; + } + + // ~/foo or ./foo + if (val.length() >= 2 && (val[0] == '~' || val[0] == '.') && val[1] == '/') + break; + + // ../foo + if (val.length() >= 3 && val[0] == '.' && val[1] == '.' && val[2] == '/') + break; + + if (Itm->Parent->Value.end()[-1] != '/') + val.insert(0, "/"); + + val.insert(0, Itm->Parent->Value); + Itm = Itm->Parent; + } + result.append(val); + } + return flNormalize(result); +} + /*}}}*/ +// Configuration::FindDir - Find a directory name /*{{{*/ +// --------------------------------------------------------------------- +/* This is like findfile execept the result is terminated in a / */ +string Configuration::FindDir(const char *Name,const char *Default) const +{ + checkFindConfigOptionType(Name, ConfigType::DIR); + string Res = FindFile(Name,Default); + if (Res.end()[-1] != '/') + { + size_t const found = Res.rfind("/dev/null"); + if (found != string::npos && found == Res.size() - 9) + return Res; // /dev/null returning + return Res + '/'; + } + return Res; +} + /*}}}*/ +// Configuration::FindVector - Find a vector of values /*{{{*/ +// --------------------------------------------------------------------- +/* Returns a vector of config values under the given item */ +vector<string> Configuration::FindVector(const char *Name, std::string const &Default, bool const Keys) const +{ + checkFindConfigOptionType(Name, ConfigType::LIST); + vector<string> Vec; + const Item *Top = Lookup(Name); + if (Top == NULL) + return VectorizeString(Default, ','); + + if (Top->Value.empty() == false) + return VectorizeString(Top->Value, ','); + + Item *I = Top->Child; + while(I != NULL) + { + Vec.push_back(Keys ? I->Tag : I->Value); + I = I->Next; + } + if (Vec.empty() == true) + return VectorizeString(Default, ','); + + return Vec; +} + /*}}}*/ +// Configuration::FindI - Find an integer value /*{{{*/ +// --------------------------------------------------------------------- +/* */ +int Configuration::FindI(const char *Name,int const &Default) const +{ + checkFindConfigOptionType(Name, ConfigType::INT); + const Item *Itm = Lookup(Name); + if (Itm == 0 || Itm->Value.empty() == true) + return Default; + + char *End; + int Res = strtol(Itm->Value.c_str(),&End,0); + if (End == Itm->Value.c_str()) + return Default; + + return Res; +} + /*}}}*/ +// Configuration::FindB - Find a boolean type /*{{{*/ +// --------------------------------------------------------------------- +/* */ +bool Configuration::FindB(const char *Name,bool const &Default) const +{ + checkFindConfigOptionType(Name, ConfigType::BOOL); + const Item *Itm = Lookup(Name); + if (Itm == 0 || Itm->Value.empty() == true) + return Default; + + return StringToBool(Itm->Value,Default); +} + /*}}}*/ +// Configuration::FindAny - Find an arbitrary type /*{{{*/ +// --------------------------------------------------------------------- +/* a key suffix of /f, /d, /b or /i calls Find{File,Dir,B,I} */ +string Configuration::FindAny(const char *Name,const char *Default) const +{ + string key = Name; + char type = 0; + + if (key.size() > 2 && key.end()[-2] == '/') + { + type = key.end()[-1]; + key.resize(key.size() - 2); + } + + switch (type) + { + // file + case 'f': + return FindFile(key.c_str(), Default); + + // directory + case 'd': + return FindDir(key.c_str(), Default); + + // bool + case 'b': + return FindB(key, Default) ? "true" : "false"; + + // int + case 'i': + { + char buf[16]; + snprintf(buf, sizeof(buf)-1, "%d", FindI(key, Default ? atoi(Default) : 0 )); + return buf; + } + } + + // fallback + return Find(Name, Default); +} + /*}}}*/ +// Configuration::CndSet - Conditional Set a value /*{{{*/ +// --------------------------------------------------------------------- +/* This will not overwrite */ +void Configuration::CndSet(const char *Name,const string &Value) +{ + Item *Itm = Lookup(Name,true); + if (Itm == 0) + return; + if (Itm->Value.empty() == true) + Itm->Value = Value; +} + /*}}}*/ +// Configuration::Set - Set an integer value /*{{{*/ +// --------------------------------------------------------------------- +/* */ +void Configuration::CndSet(const char *Name,int const Value) +{ + Item *Itm = Lookup(Name,true); + if (Itm == 0 || Itm->Value.empty() == false) + return; + char S[300]; + snprintf(S,sizeof(S),"%i",Value); + Itm->Value = S; +} + /*}}}*/ +// Configuration::Set - Set a value /*{{{*/ +// --------------------------------------------------------------------- +/* */ +void Configuration::Set(const char *Name,const string &Value) +{ + Item *Itm = Lookup(Name,true); + if (Itm == 0) + return; + Itm->Value = Value; +} + /*}}}*/ +// Configuration::Set - Set an integer value /*{{{*/ +// --------------------------------------------------------------------- +/* */ +void Configuration::Set(const char *Name,int const &Value) +{ + Item *Itm = Lookup(Name,true); + if (Itm == 0) + return; + char S[300]; + snprintf(S,sizeof(S),"%i",Value); + Itm->Value = S; +} + /*}}}*/ +// Configuration::Clear - Clear an single value from a list /*{{{*/ +// --------------------------------------------------------------------- +/* */ +void Configuration::Clear(string const &Name, int const &Value) +{ + char S[300]; + snprintf(S,sizeof(S),"%i",Value); + Clear(Name, S); +} + /*}}}*/ +// Configuration::Clear - Clear an single value from a list /*{{{*/ +// --------------------------------------------------------------------- +/* */ +void Configuration::Clear(string const &Name, string const &Value) +{ + Item *Top = Lookup(Name.c_str(),false); + if (Top == 0 || Top->Child == 0) + return; + + Item *Tmp, *Prev, *I; + Prev = I = Top->Child; + + while(I != NULL) + { + if(I->Value == Value) + { + Tmp = I; + // was first element, point parent to new first element + if(Top->Child == Tmp) + Top->Child = I->Next; + I = I->Next; + Prev->Next = I; + delete Tmp; + } else { + Prev = I; + I = I->Next; + } + } + +} + /*}}}*/ +// Configuration::Clear - Clear everything /*{{{*/ +// --------------------------------------------------------------------- +void Configuration::Clear() +{ + const Configuration::Item *Top = Tree(0); + while( Top != 0 ) + { + Clear(Top->FullTag()); + Top = Top->Next; + } +} + /*}}}*/ +// Configuration::Clear - Clear an entire tree /*{{{*/ +// --------------------------------------------------------------------- +/* */ +void Configuration::Clear(string const &Name) +{ + Item *Top = Lookup(Name.c_str(),false); + if (Top == 0) + return; + + Top->Value.clear(); + Item *Stop = Top; + Top = Top->Child; + Stop->Child = 0; + for (; Top != 0;) + { + if (Top->Child != 0) + { + Top = Top->Child; + continue; + } + + while (Top != 0 && Top->Next == 0) + { + Item *Tmp = Top; + Top = Top->Parent; + delete Tmp; + + if (Top == Stop) + return; + } + + Item *Tmp = Top; + if (Top != 0) + Top = Top->Next; + delete Tmp; + } +} + /*}}}*/ +void Configuration::MoveSubTree(char const * const OldRootName, char const * const NewRootName)/*{{{*/ +{ + // prevent NewRoot being a subtree of OldRoot + if (OldRootName == nullptr) + return; + if (NewRootName != nullptr) + { + if (strcmp(OldRootName, NewRootName) == 0) + return; + std::string const oldroot = std::string(OldRootName) + "::"; + if (strcasestr(NewRootName, oldroot.c_str()) != NULL) + return; + } + + Item * Top; + Item const * const OldRoot = Top = Lookup(OldRootName, false); + if (Top == nullptr) + return; + std::string NewRoot; + if (NewRootName != nullptr) + NewRoot.append(NewRootName).append("::"); + + Top->Value.clear(); + Item * const Stop = Top; + Top = Top->Child; + Stop->Child = 0; + for (; Top != 0;) + { + if (Top->Child != 0) + { + Top = Top->Child; + continue; + } + + while (Top != 0 && Top->Next == 0) + { + Set(NewRoot + Top->FullTag(OldRoot), Top->Value); + Item const * const Tmp = Top; + Top = Top->Parent; + delete Tmp; + + if (Top == Stop) + return; + } + + Set(NewRoot + Top->FullTag(OldRoot), Top->Value); + Item const * const Tmp = Top; + if (Top != 0) + Top = Top->Next; + delete Tmp; + } +} + /*}}}*/ +// Configuration::Exists - Returns true if the Name exists /*{{{*/ +// --------------------------------------------------------------------- +/* */ +bool Configuration::Exists(const char *Name) const +{ + const Item *Itm = Lookup(Name); + if (Itm == 0) + return false; + return true; +} + /*}}}*/ +// Configuration::ExistsAny - Returns true if the Name, possibly /*{{{*/ +// --------------------------------------------------------------------- +/* qualified by /[fdbi] exists */ +bool Configuration::ExistsAny(const char *Name) const +{ + string key = Name; + + if (key.size() > 2 && key.end()[-2] == '/') + { + if (key.find_first_of("fdbi",key.size()-1) < key.size()) + { + key.resize(key.size() - 2); + if (Exists(key.c_str())) + return true; + } + else + { + _error->Warning(_("Unrecognized type abbreviation: '%c'"), key.end()[-3]); + } + } + return Exists(Name); +} + /*}}}*/ +// Configuration::Dump - Dump the config /*{{{*/ +// --------------------------------------------------------------------- +/* Dump the entire configuration space */ +void Configuration::Dump(ostream& str) +{ + Dump(str, NULL, "%F \"%v\";\n", true); +} +void Configuration::Dump(ostream& str, char const * const root, + char const * const formatstr, bool const emptyValue) +{ + const Configuration::Item* Top = Tree(root); + if (Top == 0) + return; + const Configuration::Item* const Root = (root == NULL) ? NULL : Top; + std::vector<std::string> const format = VectorizeString(formatstr, '%'); + + /* Write out all of the configuration directives by walking the + configuration tree */ + do { + if (emptyValue == true || Top->Value.empty() == emptyValue) + { + std::vector<std::string>::const_iterator f = format.begin(); + str << *f; + for (++f; f != format.end(); ++f) + { + if (f->empty() == true) + { + ++f; + str << '%' << *f; + continue; + } + char const type = (*f)[0]; + if (type == 'f') + str << Top->FullTag(); + else if (type == 't') + str << Top->Tag; + else if (type == 'v') + str << Top->Value; + else if (type == 'F') + str << QuoteString(Top->FullTag(), "=\"\n"); + else if (type == 'T') + str << QuoteString(Top->Tag, "=\"\n"); + else if (type == 'V') + str << QuoteString(Top->Value, "=\"\n"); + else if (type == 'n') + str << "\n"; + else if (type == 'N') + str << "\t"; + else + str << '%' << type; + str << f->c_str() + 1; + } + } + + if (Top->Child != 0) + { + Top = Top->Child; + continue; + } + + while (Top != 0 && Top->Next == 0) + Top = Top->Parent; + if (Top != 0) + Top = Top->Next; + + if (Root != NULL) + { + const Configuration::Item* I = Top; + while(I != 0) + { + if (I == Root) + break; + else + I = I->Parent; + } + if (I == 0) + break; + } + } while (Top != 0); +} + /*}}}*/ + +// Configuration::Item::FullTag - Return the fully scoped tag /*{{{*/ +// --------------------------------------------------------------------- +/* Stop sets an optional max recursion depth if this item is being viewed as + part of a sub tree. */ +string Configuration::Item::FullTag(const Item *Stop) const +{ + if (Parent == 0 || Parent->Parent == 0 || Parent == Stop) + return Tag; + return Parent->FullTag(Stop) + "::" + Tag; +} + /*}}}*/ + +// ReadConfigFile - Read a configuration file /*{{{*/ +// --------------------------------------------------------------------- +/* The configuration format is very much like the named.conf format + used in bind8, in fact this routine can parse most named.conf files. + Sectional config files are like bind's named.conf where there are + sections like 'zone "foo.org" { .. };' This causes each section to be + added in with a tag like "zone::foo.org" instead of being split + tag/value. AsSectional enables Sectional parsing.*/ +static void leaveCurrentScope(std::stack<std::string> &Stack, std::string &ParentTag) +{ + if (Stack.empty()) + ParentTag.clear(); + else + { + ParentTag = Stack.top(); + Stack.pop(); + } +} +bool ReadConfigFile(Configuration &Conf,const string &FName,bool const &AsSectional, + unsigned const &Depth) +{ + // Open the stream for reading + FileFd F; + if (OpenConfigurationFileFd(FName, F) == false) + return false; + + string LineBuffer; + std::stack<std::string> Stack; + + // Parser state + string ParentTag; + + int CurLine = 0; + bool InComment = false; + while (F.Eof() == false) + { + // The raw input line. + std::string Input; + if (F.ReadLine(Input) == false) + Input.clear(); + // The input line with comments stripped. + std::string Fragment; + + // Expand tabs in the input line and remove leading and trailing whitespace. + Input = APT::String::Strip(SubstVar(Input, "\t", " ")); + CurLine++; + + // Now strip comments; if the whole line is contained in a + // comment, skip this line. + APT::StringView Line{Input.data(), Input.size()}; + + // continued Multi line comment + if (InComment) + { + size_t end = Line.find("*/"); + if (end != APT::StringView::npos) + { + Line.remove_prefix(end + 2); + InComment = false; + } + else + continue; + } + + // Discard single line comments + { + size_t start = 0; + while ((start = Line.find("//", start)) != APT::StringView::npos) + { + if (std::count(Line.begin(), Line.begin() + start, '"') % 2 != 0) + { + ++start; + continue; + } + Line.remove_suffix(Line.length() - start); + break; + } + using APT::operator""_sv; + constexpr std::array<APT::StringView, 3> magicComments { "clear"_sv, "include"_sv, "x-apt-configure-index"_sv }; + start = 0; + while ((start = Line.find('#', start)) != APT::StringView::npos) + { + if (std::count(Line.begin(), Line.begin() + start, '"') % 2 != 0 || + std::any_of(magicComments.begin(), magicComments.end(), [&](auto const m) { return Line.compare(start+1, m.length(), m) == 0; })) + { + ++start; + continue; + } + Line.remove_suffix(Line.length() - start); + break; + } + } + + // Look for multi line comments and build up the + // fragment. + Fragment.reserve(Line.length()); + { + size_t start = 0; + while ((start = Line.find("/*", start)) != APT::StringView::npos) + { + if (std::count(Line.begin(), Line.begin() + start, '"') % 2 != 0) + { + start += 2; + continue; + } + Fragment.append(Line.data(), start); + auto const end = Line.find("*/", start + 2); + if (end == APT::StringView::npos) + { + Line.clear(); + InComment = true; + break; + } + else + Line.remove_prefix(end + 2); + start = 0; + } + if (not Line.empty()) + Fragment.append(Line.data(), Line.length()); + } + + // Skip blank lines. + if (Fragment.empty()) + continue; + + // The line has actual content; interpret what it means. + bool InQuote = false; + auto Start = Fragment.cbegin(); + auto End = Fragment.cend(); + for (std::string::const_iterator I = Start; + I != End; ++I) + { + if (*I == '"') + InQuote = !InQuote; + + if (InQuote == false && (*I == '{' || *I == ';' || *I == '}')) + { + // Put the last fragment into the buffer + std::string::const_iterator NonWhitespaceStart = Start; + std::string::const_iterator NonWhitespaceStop = I; + for (; NonWhitespaceStart != I && isspace(*NonWhitespaceStart) != 0; ++NonWhitespaceStart) + ; + for (; NonWhitespaceStop != NonWhitespaceStart && isspace(NonWhitespaceStop[-1]) != 0; --NonWhitespaceStop) + ; + if (LineBuffer.empty() == false && NonWhitespaceStop - NonWhitespaceStart != 0) + LineBuffer += ' '; + LineBuffer += string(NonWhitespaceStart, NonWhitespaceStop); + + // Drop this from the input string, saving the character + // that terminated the construct we just closed. (i.e., a + // brace or a semicolon) + char TermChar = *I; + Start = I + 1; + + // Syntax Error + if (TermChar == '{' && LineBuffer.empty() == true) + return _error->Error(_("Syntax error %s:%u: Block starts with no name."),FName.c_str(),CurLine); + + // No string on this line + if (LineBuffer.empty() == true) + { + if (TermChar == '}') + leaveCurrentScope(Stack, ParentTag); + continue; + } + + // Parse off the tag + string Tag; + const char *Pos = LineBuffer.c_str(); + if (ParseQuoteWord(Pos,Tag) == false) + return _error->Error(_("Syntax error %s:%u: Malformed tag"),FName.c_str(),CurLine); + + // Parse off the word + string Word; + bool NoWord = false; + if (ParseCWord(Pos,Word) == false && + ParseQuoteWord(Pos,Word) == false) + { + if (TermChar != '{') + { + Word = Tag; + Tag = ""; + } + else + NoWord = true; + } + if (strlen(Pos) != 0) + return _error->Error(_("Syntax error %s:%u: Extra junk after value"),FName.c_str(),CurLine); + + // Go down a level + if (TermChar == '{') + { + Stack.push(ParentTag); + + /* Make sectional tags incorporate the section into the + tag string */ + if (AsSectional == true && Word.empty() == false) + { + Tag.append("::").append(Word); + Word.clear(); + } + + if (ParentTag.empty() == true) + ParentTag = Tag; + else + ParentTag.append("::").append(Tag); + Tag.clear(); + } + + // Generate the item name + string Item; + if (ParentTag.empty() == true) + Item = Tag; + else + { + if (TermChar != '{' || Tag.empty() == false) + Item = ParentTag + "::" + Tag; + else + Item = ParentTag; + } + + // Specials + if (Tag.length() >= 1 && Tag[0] == '#') + { + if (ParentTag.empty() == false) + return _error->Error(_("Syntax error %s:%u: Directives can only be done at the top level"),FName.c_str(),CurLine); + Tag.erase(Tag.begin()); + if (Tag == "clear") + Conf.Clear(Word); + else if (Tag == "include") + { + if (Depth > 10) + return _error->Error(_("Syntax error %s:%u: Too many nested includes"),FName.c_str(),CurLine); + if (Word.length() > 2 && Word.end()[-1] == '/') + { + if (ReadConfigDir(Conf,Word,AsSectional,Depth+1) == false) + return _error->Error(_("Syntax error %s:%u: Included from here"),FName.c_str(),CurLine); + } + else + { + if (ReadConfigFile(Conf,Word,AsSectional,Depth+1) == false) + return _error->Error(_("Syntax error %s:%u: Included from here"),FName.c_str(),CurLine); + } + } + else if (Tag == "x-apt-configure-index") + { + if (LoadConfigurationIndex(Word) == false) + return _error->Warning("Loading the configure index %s in file %s:%u failed!", Word.c_str(), FName.c_str(), CurLine); + } + else + return _error->Error(_("Syntax error %s:%u: Unsupported directive '%s'"),FName.c_str(),CurLine,Tag.c_str()); + } + else if (Tag.empty() == true && NoWord == false && Word == "#clear") + return _error->Error(_("Syntax error %s:%u: clear directive requires an option tree as argument"),FName.c_str(),CurLine); + else + { + // Set the item in the configuration class + if (NoWord == false) + Conf.Set(Item,Word); + } + + // Empty the buffer + LineBuffer.clear(); + + // Move up a tag, but only if there is no bit to parse + if (TermChar == '}') + leaveCurrentScope(Stack, ParentTag); + } + } + + // Store the remaining text, if any, in the current line buffer. + + // NB: could change this to use string-based operations; I'm + // using strstrip now to ensure backwards compatibility. + // -- dburrows 2008-04-01 + { + char *Buffer = new char[End - Start + 1]; + try + { + std::copy(Start, End, Buffer); + Buffer[End - Start] = '\0'; + + const char *Stripd = _strstrip(Buffer); + if (*Stripd != 0 && LineBuffer.empty() == false) + LineBuffer += " "; + LineBuffer += Stripd; + } + catch(...) + { + delete[] Buffer; + throw; + } + delete[] Buffer; + } + } + + if (LineBuffer.empty() == false) + return _error->Error(_("Syntax error %s:%u: Extra junk at end of file"),FName.c_str(),CurLine); + return true; +} + /*}}}*/ +// ReadConfigDir - Read a directory of config files /*{{{*/ +// --------------------------------------------------------------------- +/* */ +bool ReadConfigDir(Configuration &Conf,const string &Dir, + bool const &AsSectional, unsigned const &Depth) +{ + _error->PushToStack(); + auto const files = GetListOfFilesInDir(Dir, "conf", true, true); + auto const successfulList = not _error->PendingError(); + _error->MergeWithStack(); + return std::accumulate(files.cbegin(), files.cend(), true, [&](bool good, auto const &file) { + return ReadConfigFile(Conf, file, AsSectional, Depth) && good; + }) && successfulList; +} + /*}}}*/ +// MatchAgainstConfig Constructor /*{{{*/ +Configuration::MatchAgainstConfig::MatchAgainstConfig(char const * Config) +{ + std::vector<std::string> const strings = _config->FindVector(Config); + for (std::vector<std::string>::const_iterator s = strings.begin(); + s != strings.end(); ++s) + { + regex_t *p = new regex_t; + if (regcomp(p, s->c_str(), REG_EXTENDED | REG_ICASE | REG_NOSUB) == 0) + patterns.push_back(p); + else + { + regfree(p); + delete p; + _error->Warning("Invalid regular expression '%s' in configuration " + "option '%s' will be ignored.", + s->c_str(), Config); + continue; + } + } + if (strings.empty() == true) + patterns.push_back(NULL); +} + /*}}}*/ +// MatchAgainstConfig Destructor /*{{{*/ +Configuration::MatchAgainstConfig::~MatchAgainstConfig() +{ + clearPatterns(); +} +void Configuration::MatchAgainstConfig::clearPatterns() +{ + for(std::vector<regex_t *>::const_iterator p = patterns.begin(); + p != patterns.end(); ++p) + { + if (*p == NULL) continue; + regfree(*p); + delete *p; + } + patterns.clear(); +} + /*}}}*/ +// MatchAgainstConfig::Match - returns true if a pattern matches /*{{{*/ +bool Configuration::MatchAgainstConfig::Match(char const * str) const +{ + for(std::vector<regex_t *>::const_iterator p = patterns.begin(); + p != patterns.end(); ++p) + if (*p != NULL && regexec(*p, str, 0, 0, 0) == 0) + return true; + + return false; +} + /*}}}*/ diff --git a/apt-pkg/contrib/configuration.h b/apt-pkg/contrib/configuration.h new file mode 100644 index 0000000..6ebf28d --- /dev/null +++ b/apt-pkg/contrib/configuration.h @@ -0,0 +1,152 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + Configuration Class + + This class provides a configuration file and command line parser + for a tree-oriented configuration environment. All runtime configuration + is stored in here. + + Each configuration name is given as a fully scoped string such as + Foo::Bar + And has associated with it a text string. The Configuration class only + provides storage and lookup for this tree, other classes provide + configuration file formats (and parsers/emitters if needed). + + Most things can get by quite happily with, + cout << _config->Find("Foo::Bar") << endl; + + A special extension, support for ordered lists is provided by using the + special syntax, "block::list::" the trailing :: designates the + item as a list. To access the list you must use the tree function on + "block::list". + + ##################################################################### */ + /*}}}*/ +#ifndef PKGLIB_CONFIGURATION_H +#define PKGLIB_CONFIGURATION_H + +#include <regex.h> + +#include <iostream> +#include <string> +#include <vector> + +#include <apt-pkg/macros.h> + + +class APT_PUBLIC Configuration +{ + public: + + struct Item + { + std::string Value; + std::string Tag; + Item *Parent; + Item *Child; + Item *Next; + + std::string FullTag(const Item *Stop = 0) const; + + Item() : Parent(0), Child(0), Next(0) {}; + }; + + private: + + Item *Root; + bool ToFree; + + Item *Lookup(Item *Head,const char *S,unsigned long const &Len,bool const &Create); + Item *Lookup(const char *Name,const bool &Create); + inline const Item *Lookup(const char *Name) const + { + return const_cast<Configuration *>(this)->Lookup(Name,false); + } + + public: + + std::string Find(const char *Name,const char *Default = 0) const; + std::string Find(std::string const &Name,const char *Default = 0) const {return Find(Name.c_str(),Default);}; + std::string Find(std::string const &Name, std::string const &Default) const {return Find(Name.c_str(),Default.c_str());}; + std::string FindFile(const char *Name,const char *Default = 0) const; + std::string FindDir(const char *Name,const char *Default = 0) const; + /** return a list of child options + * + * Options like Acquire::Languages are handled as lists which + * can be overridden and have a default. For the later two a comma + * separated list of values is supported. + * + * \param Name of the parent node + * \param Default list of values separated by commas */ + std::vector<std::string> FindVector(const char *Name, std::string const &Default = "", bool const Keys = false) const; + std::vector<std::string> FindVector(std::string const &Name, std::string const &Default = "", bool const Keys = false) const { return FindVector(Name.c_str(), Default, Keys); }; + + int FindI(const char *Name,int const &Default = 0) const; + int FindI(std::string const &Name,int const &Default = 0) const {return FindI(Name.c_str(),Default);}; + bool FindB(const char *Name,bool const &Default = false) const; + bool FindB(std::string const &Name,bool const &Default = false) const {return FindB(Name.c_str(),Default);}; + std::string FindAny(const char *Name,const char *Default = 0) const; + + inline void Set(const std::string &Name,const std::string &Value) {Set(Name.c_str(),Value);}; + void CndSet(const char *Name,const std::string &Value); + void CndSet(const char *Name,const int Value); + void Set(const char *Name,const std::string &Value); + void Set(const char *Name,const int &Value); + + inline bool Exists(const std::string &Name) const {return Exists(Name.c_str());}; + bool Exists(const char *Name) const; + bool ExistsAny(const char *Name) const; + + void MoveSubTree(char const * const OldRoot, char const * const NewRoot); + + // clear a whole tree + void Clear(const std::string &Name); + void Clear(); + + // remove a certain value from a list (e.g. the list of "APT::Keep-Fds") + void Clear(std::string const &List, std::string const &Value); + void Clear(std::string const &List, int const &Value); + + inline const Item *Tree(const char *Name) const {return Lookup(Name);}; + + inline void Dump() { Dump(std::clog); }; + void Dump(std::ostream& str); + void Dump(std::ostream& str, char const * const root, + char const * const format, bool const emptyValue); + + explicit Configuration(const Item *Root); + Configuration(); + ~Configuration(); + + /** \brief match a string against a configurable list of patterns */ + class MatchAgainstConfig + { + std::vector<regex_t *> patterns; + APT_HIDDEN void clearPatterns(); + + public: + explicit MatchAgainstConfig(char const * Config); + virtual ~MatchAgainstConfig(); + + /** \brief Returns \b true for a string matching one of the patterns */ + bool Match(char const * str) const; + bool Match(std::string const &str) const { return Match(str.c_str()); }; + + /** \brief returns if the matcher setup was successful */ + bool wasConstructedSuccessfully() const { return patterns.empty() == false; } + }; +}; + +APT_PUBLIC extern Configuration *_config; + +APT_PUBLIC bool ReadConfigFile(Configuration &Conf,const std::string &FName, + bool const &AsSectional = false, + unsigned const &Depth = 0); + +APT_PUBLIC bool ReadConfigDir(Configuration &Conf,const std::string &Dir, + bool const &AsSectional = false, + unsigned const &Depth = 0); + +#endif diff --git a/apt-pkg/contrib/error.cc b/apt-pkg/contrib/error.cc new file mode 100644 index 0000000..dd28af1 --- /dev/null +++ b/apt-pkg/contrib/error.cc @@ -0,0 +1,336 @@ +// -*- mode: cpp; mode: fold -*- +// SPDX-License-Identifier: GPL-2.0+ +// Description /*{{{*/ +/* ###################################################################### + + Global Error Class - Global error mechanism + + We use a simple STL vector to store each error record. A PendingFlag + is kept which indicates when the vector contains a Sever error. + + This file had this historic note, but now includes further changes + under the GPL-2.0+: + + This source is placed in the Public Domain, do with it what you will + It was originally written by Jason Gunthorpe. + + ##################################################################### */ + /*}}}*/ +// Include Files /*{{{*/ +#include <config.h> + +#include <apt-pkg/configuration.h> +#include <apt-pkg/error.h> + +#include <algorithm> +#include <cstring> +#include <iostream> +#include <list> +#include <string> +#include <errno.h> +#include <stdarg.h> +#include <stddef.h> +#include <stdio.h> +#include <stdlib.h> +#include <unistd.h> + + /*}}}*/ + +// Global Error Object /*{{{*/ +GlobalError *_GetErrorObj() +{ + static thread_local GlobalError Obj; + return &Obj; +} + /*}}}*/ +// GlobalError::GlobalError - Constructor /*{{{*/ +GlobalError::GlobalError() : PendingFlag(false) {} + /*}}}*/ +// GlobalError::FatalE, Errno, WarningE, NoticeE and DebugE - Add to the list/*{{{*/ +#define GEMessage(NAME, TYPE) \ +bool GlobalError::NAME (const char *Function, const char *Description,...) { \ + va_list args; \ + size_t msgSize = 400; \ + int const errsv = errno; \ + bool retry; \ + do { \ + va_start(args,Description); \ + retry = InsertErrno(TYPE, Function, Description, args, errsv, msgSize); \ + va_end(args); \ + } while (retry); \ + return false; \ +} +GEMessage(FatalE, FATAL) +GEMessage(Errno, ERROR) +GEMessage(WarningE, WARNING) +GEMessage(NoticeE, NOTICE) +GEMessage(DebugE, DEBUG) +#undef GEMessage + /*}}}*/ +// GlobalError::InsertErrno - Get part of the errortype string from errno/*{{{*/ +bool GlobalError::InsertErrno(MsgType const &type, const char *Function, + const char *Description,...) { + va_list args; + size_t msgSize = 400; + int const errsv = errno; + bool retry; + do { + va_start(args,Description); + retry = InsertErrno(type, Function, Description, args, errsv, msgSize); + va_end(args); + } while (retry); + return false; +} + /*}}}*/ +// GlobalError::InsertErrno - formats an error message with the errno /*{{{*/ +bool GlobalError::InsertErrno(MsgType type, const char* Function, + const char* Description, va_list &args, + int const errsv, size_t &msgSize) { + char* S = (char*) malloc(msgSize); + int const n = snprintf(S, msgSize, "%s - %s (%i: %s)", Description, + Function, errsv, strerror(errsv)); + if (n > -1 && ((unsigned int) n) < msgSize); + else { + if (n > -1) + msgSize = n + 1; + else + msgSize *= 2; + free(S); + return true; + } + + bool const geins = Insert(type, S, args, msgSize); + free(S); + return geins; +} + /*}}}*/ +// GlobalError::Fatal, Error, Warning, Notice and Debug - Add to the list/*{{{*/ +#define GEMessage(NAME, TYPE) \ +bool GlobalError::NAME (const char *Description,...) { \ + va_list args; \ + size_t msgSize = 400; \ + bool retry; \ + do { \ + va_start(args,Description); \ + retry = Insert(TYPE, Description, args, msgSize); \ + va_end(args); \ + } while (retry); \ + return false; \ +} +GEMessage(Fatal, FATAL) +GEMessage(Error, ERROR) +GEMessage(Warning, WARNING) +GEMessage(Notice, NOTICE) +GEMessage(Debug, DEBUG) +#undef GEMessage + /*}}}*/ +// GlobalError::Insert - Add a errotype message to the list /*{{{*/ +bool GlobalError::Insert(MsgType const &type, const char *Description,...) +{ + va_list args; + size_t msgSize = 400; + bool retry; + do { + va_start(args,Description); + retry = Insert(type, Description, args, msgSize); + va_end(args); + } while (retry); + return false; +} + /*}}}*/ +// GlobalError::Insert - Insert a new item at the end /*{{{*/ +bool GlobalError::Insert(MsgType type, const char* Description, + va_list &args, size_t &msgSize) { + char* S = (char*) malloc(msgSize); + int const n = vsnprintf(S, msgSize, Description, args); + if (n > -1 && ((unsigned int) n) < msgSize); + else { + if (n > -1) + msgSize = n + 1; + else + msgSize *= 2; + free(S); + return true; + } + + Item const m(S, type); + Messages.push_back(m); + + if (type == ERROR || type == FATAL) + PendingFlag = true; + + if (type == FATAL || type == DEBUG) + std::clog << m << std::endl; + + free(S); + return false; +} + /*}}}*/ +// GlobalError::PopMessage - Pulls a single message out /*{{{*/ +bool GlobalError::PopMessage(std::string &Text) { + if (Messages.empty() == true) + return false; + + Item const msg = Messages.front(); + Messages.pop_front(); + + bool const Ret = (msg.Type == ERROR || msg.Type == FATAL); + Text = msg.Text; + if (PendingFlag == false || Ret == false) + return Ret; + + // check if another error message is pending + for (std::list<Item>::const_iterator m = Messages.begin(); + m != Messages.end(); ++m) + if (m->Type == ERROR || m->Type == FATAL) + return Ret; + + PendingFlag = false; + return Ret; +} + /*}}}*/ +// GlobalError::DumpErrors - Dump all of the errors/warns to cerr /*{{{*/ +void GlobalError::DumpErrors(std::ostream &out, MsgType const &threshold, + bool const &mergeStack) { + if (mergeStack == true) + for (std::list<MsgStack>::const_reverse_iterator s = Stacks.rbegin(); + s != Stacks.rend(); ++s) + std::copy(s->Messages.begin(), s->Messages.end(), std::front_inserter(Messages)); + + std::for_each(Messages.begin(), Messages.end(), [&threshold, &out](Item const &m) { + if (m.Type >= threshold) + out << m << std::endl; + }); + + Discard(); +} + /*}}}*/ +// GlobalError::Discard - Discard /*{{{*/ +void GlobalError::Discard() { + Messages.clear(); + PendingFlag = false; +} + /*}}}*/ +// GlobalError::empty - does our error list include anything? /*{{{*/ +bool GlobalError::empty(MsgType const &threshold) const { + if (PendingFlag == true) + return false; + + if (Messages.empty() == true) + return true; + + return std::find_if(Messages.begin(), Messages.end(), [&threshold](Item const &m) { + return m.Type >= threshold; + }) == Messages.end(); +} + /*}}}*/ +// GlobalError::PushToStack /*{{{*/ +void GlobalError::PushToStack() { + Stacks.emplace_back(Messages, PendingFlag); + Discard(); +} + /*}}}*/ +// GlobalError::RevertToStack /*{{{*/ +void GlobalError::RevertToStack() { + Discard(); + MsgStack pack = Stacks.back(); + Messages = pack.Messages; + PendingFlag = pack.PendingFlag; + Stacks.pop_back(); +} + /*}}}*/ +// GlobalError::MergeWithStack /*{{{*/ +void GlobalError::MergeWithStack() { + MsgStack pack = Stacks.back(); + Messages.splice(Messages.begin(), pack.Messages); + PendingFlag = PendingFlag || pack.PendingFlag; + Stacks.pop_back(); +} + /*}}}*/ + +// GlobalError::Item::operator<< /*{{{*/ +APT_HIDDEN std::ostream &operator<<(std::ostream &out, GlobalError::Item i) +{ + static constexpr auto COLOR_RESET = "\033[0m"; + static constexpr auto COLOR_NOTICE = "\033[33m"; // normal yellow + static constexpr auto COLOR_WARN = "\033[1;33m"; // bold yellow + static constexpr auto COLOR_ERROR = "\033[1;31m"; // bold red + + bool use_color = _config->FindB("APT::Color", false); + + if (use_color) + { + switch (i.Type) + { + case GlobalError::FATAL: + case GlobalError::ERROR: + out << COLOR_ERROR; + break; + case GlobalError::WARNING: + out << COLOR_WARN; + break; + case GlobalError::NOTICE: + out << COLOR_NOTICE; + break; + default: + break; + } + } + + switch (i.Type) + { + case GlobalError::FATAL: + case GlobalError::ERROR: + out << 'E'; + break; + case GlobalError::WARNING: + out << 'W'; + break; + case GlobalError::NOTICE: + out << 'N'; + break; + case GlobalError::DEBUG: + out << 'D'; + break; + } + out << ": "; + + if (use_color) + { + switch (i.Type) + { + case GlobalError::FATAL: + case GlobalError::ERROR: + case GlobalError::WARNING: + case GlobalError::NOTICE: + out << COLOR_RESET; + break; + default: + break; + } + } + + std::string::size_type line_start = 0; + std::string::size_type line_end; + while ((line_end = i.Text.find_first_of("\n\r", line_start)) != std::string::npos) + { + if (line_start != 0) + out << std::endl + << " "; + out << i.Text.substr(line_start, line_end - line_start); + line_start = i.Text.find_first_not_of("\n\r", line_end + 1); + if (line_start == std::string::npos) + break; + } + if (line_start == 0) + out << i.Text; + else if (line_start != std::string::npos) + out << std::endl + << " " << i.Text.substr(line_start); + + if (use_color) + out << COLOR_RESET; + + return out; +} + /*}}}*/ diff --git a/apt-pkg/contrib/error.h b/apt-pkg/contrib/error.h new file mode 100644 index 0000000..dd50b9c --- /dev/null +++ b/apt-pkg/contrib/error.h @@ -0,0 +1,349 @@ +// -*- mode: cpp; mode: fold -*- +// SPDX-License-Identifier: GPL-2.0+ +// Description /*{{{*/ +/* ###################################################################### + + Global Error Class - Global error mechanism + + This class has a single global instance. When a function needs to + generate an error condition, such as a read error, it calls a member + in this class to add the error to a stack of errors. + + By using a stack the problem with a scheme like errno is removed and + it allows a very detailed account of what went wrong to be transmitted + to the UI for display. (Errno has problems because each function sets + errno to 0 if it didn't have an error thus eraseing erno in the process + of cleanup) + + Several predefined error generators are provided to handle common + things like errno. The general idea is that all methods return a bool. + If the bool is true then things are OK, if it is false then things + should start being undone and the stack should unwind under program + control. + + A Warning should not force the return of false. Things did not fail, but + they might have had unexpected problems. Errors are stored in a FIFO + so Pop will return the first item.. + + I have some thoughts about extending this into a more general UI<-> + Engine interface, ie allowing the Engine to say 'The disk is full' in + a dialog that says 'Panic' and 'Retry'.. The error generator functions + like errno, Warning and Error return false always so this is normal: + if (open(..)) + return _error->Errno(..); + + This file had this historic note, but now includes further changes + under the GPL-2.0+: + + This source is placed in the Public Domain, do with it what you will + It was originally written by Jason Gunthorpe. + + ##################################################################### */ + /*}}}*/ +#ifndef PKGLIB_ERROR_H +#define PKGLIB_ERROR_H + +#include <apt-pkg/macros.h> + +#include <iostream> +#include <list> +#include <string> + +#include <stdarg.h> +#include <stddef.h> + +class APT_PUBLIC GlobalError /*{{{*/ +{ +public: /*{{{*/ + /** \brief a message can have one of following severity */ + enum MsgType { + /** \brief Message will be printed instantly as it is likely that + this error will lead to a complete crash */ + FATAL = 40, + /** \brief An error does hinder the correct execution and should be corrected */ + ERROR = 30, + /** \brief indicates problem that can lead to errors later on */ + WARNING = 20, + /** \brief deprecation warnings, old fallback behavior, … */ + NOTICE = 10, + /** \brief for developers only in areas it is hard to print something directly */ + DEBUG = 0 + }; + + /** \brief add a fatal error message with errno to the list + * + * \param Function name of the function generating the error + * \param Description format string for the error message + * + * \return \b false + */ + bool FatalE(const char *Function,const char *Description,...) APT_PRINTF(3) APT_COLD; + + /** \brief add an Error message with errno to the list + * + * \param Function name of the function generating the error + * \param Description format string for the error message + * + * \return \b false + */ + bool Errno(const char *Function,const char *Description,...) APT_PRINTF(3) APT_COLD; + + /** \brief add a warning message with errno to the list + * + * A warning should be considered less severe than an error and + * may be ignored by the client. + * + * \param Function Name of the function generates the warning. + * \param Description Format string for the warning message. + * + * \return \b false + */ + bool WarningE(const char *Function,const char *Description,...) APT_PRINTF(3) APT_COLD; + + /** \brief add a notice message with errno to the list + * + * \param Function name of the function generating the error + * \param Description format string for the error message + * + * \return \b false + */ + bool NoticeE(const char *Function,const char *Description,...) APT_PRINTF(3) APT_COLD; + + /** \brief add a debug message with errno to the list + * + * \param Function name of the function generating the error + * \param Description format string for the error message + * + * \return \b false + */ + bool DebugE(const char *Function,const char *Description,...) APT_PRINTF(3) APT_COLD; + + /** \brief adds an errno message with the given type + * + * \param type of the error message + * \param Function which failed + * \param Description of the error + */ + bool InsertErrno(MsgType const &type, const char* Function, + const char* Description,...) APT_PRINTF(4) APT_COLD; + + /** \brief adds an errno message with the given type + * + * args needs to be initialized with va_start and terminated + * with va_end by the caller. msgSize is also an out-parameter + * in case the msgSize was not enough to store the complete message. + * + * \param type of the error message + * \param Function which failed + * \param Description is the format string for args + * \param args list from a printf-like function + * \param errsv is the errno the error is for + * \param msgSize is the size of the char[] used to store message + * \return true if the message was added, false if not - the caller + * should call this method again in that case + */ + bool InsertErrno(MsgType type, const char* Function, + const char* Description, va_list &args, + int const errsv, size_t &msgSize) APT_COLD; + + /** \brief add an fatal error message to the list + * + * Most of the stuff we consider as "error" is also "fatal" for + * the user as the application will not have the expected result, + * but a fatal message here means that it gets printed directly + * to stderr in addition to adding it to the list as the error + * leads sometimes to crashes and a maybe duplicated message + * is better than "Segfault" as the only displayed text + * + * \param Description Format string for the fatal error message. + * + * \return \b false + */ + bool Fatal(const char *Description,...) APT_PRINTF(2) APT_COLD; + + /** \brief add an Error message to the list + * + * \param Description Format string for the error message. + * + * \return \b false + */ + bool Error(const char *Description,...) APT_PRINTF(2) APT_COLD; + + /** \brief add a warning message to the list + * + * A warning should be considered less severe than an error and + * may be ignored by the client. + * + * \param Description Format string for the message + * + * \return \b false + */ + bool Warning(const char *Description,...) APT_PRINTF(2) APT_COLD; + + /** \brief add a notice message to the list + * + * A notice should be considered less severe than an error or a + * warning and can be ignored by the client without further problems + * for some times, but he should consider fixing the problem. + * This error type can be used for e.g. deprecation warnings of options. + * + * \param Description Format string for the message + * + * \return \b false + */ + bool Notice(const char *Description,...) APT_PRINTF(2) APT_COLD; + + /** \brief add a debug message to the list + * + * \param Description Format string for the message + * + * \return \b false + */ + bool Debug(const char *Description,...) APT_PRINTF(2) APT_COLD; + + /** \brief adds an error message with the given type + * + * \param type of the error message + * \param Description of the error + */ + bool Insert(MsgType const &type, const char* Description,...) APT_PRINTF(3) APT_COLD; + + /** \brief adds an error message with the given type + * + * args needs to be initialized with va_start and terminated + * with va_end by the caller. msgSize is also an out-parameter + * in case the msgSize was not enough to store the complete message. + * + * \param type of the error message + * \param Description is the format string for args + * \param args list from a printf-like function + * \param msgSize is the size of the char[] used to store message + * \return true if the message was added, false if not - the caller + * should call this method again in that case + */ + bool Insert(MsgType type, const char* Description, + va_list &args, size_t &msgSize) APT_COLD; + + /** \brief is an error in the list? + * + * \return \b true if an error is included in the list, \b false otherwise + */ + inline bool PendingError() const APT_PURE {return PendingFlag;}; + + /** \brief is the list empty? + * + * Can be used to check if the current stack level doesn't include + * anything equal or more severe than a given threshold, defaulting + * to warning level for historic reasons. + * + * \param threshold minimum level considered + * + * \return \b true if the list is empty, \b false otherwise + */ + bool empty(MsgType const &threshold = WARNING) const APT_PURE; + + /** \brief returns and removes the first (or last) message in the list + * + * \param[out] Text message of the first/last item + * + * \return \b true if the message was an error, \b false otherwise + */ + bool PopMessage(std::string &Text); + + /** \brief clears the list of messages */ + void Discard(); + + /** \brief outputs the list of messages to the given stream + * + * Note that all messages are discarded, even undisplayed ones. + * + * \param[out] out output stream to write the messages in + * \param threshold minimum level considered + * \param mergeStack if true recursively dumps the entire stack + */ + void DumpErrors(std::ostream &out, MsgType const &threshold = WARNING, + bool const &mergeStack = true); + + /** \brief dumps the list of messages to std::cerr + * + * Note that all messages are discarded, also the notices + * displayed or not. + * + * \param threshold minimum level printed + */ + void inline DumpErrors(MsgType const &threshold) { + DumpErrors(std::cerr, threshold); + } + + // mvo: we do this instead of using a default parameter in the + // previous declaration to avoid a (subtle) API break for + // e.g. sigc++ and mem_fun0 + /** \brief dumps the messages of type WARNING or higher to std::cerr + * + * Note that all messages are discarded, displayed or not. + * + */ + void inline DumpErrors() { + DumpErrors(WARNING); + } + + /** \brief put the current Messages into the stack + * + * All "old" messages will be pushed into a stack to + * them later back, but for now the Message query will be + * empty and performs as no messages were present before. + * + * The stack can be as deep as you want - all stack operations + * will only operate on the last element in the stack. + */ + void PushToStack(); + + /** \brief throw away all current messages */ + void RevertToStack(); + + /** \brief merge current and stack together */ + void MergeWithStack(); + + /** \brief return the deep of the stack */ + size_t StackCount() const APT_PURE { + return Stacks.size(); + } + + GlobalError(); + /*}}}*/ +private: /*{{{*/ + struct Item { + std::string Text; + MsgType Type; + + Item(char const *Text, MsgType const &Type) : + Text(Text), Type(Type) {}; + + APT_HIDDEN friend std::ostream &operator<<(std::ostream &out, Item i); + }; + + APT_HIDDEN friend std::ostream &operator<<(std::ostream &out, Item i); + + std::list<Item> Messages; + bool PendingFlag; + + struct MsgStack { + std::list<Item> Messages; + bool const PendingFlag; + + MsgStack(std::list<Item> const &Messages, bool const &Pending) : + Messages(Messages), PendingFlag(Pending) {}; + }; + + std::list<MsgStack> Stacks; + /*}}}*/ +}; + /*}}}*/ + +// The 'extra-ansi' syntax is used to help with collisions. +APT_PUBLIC GlobalError *_GetErrorObj(); +static struct { + inline GlobalError* operator ->() { return _GetErrorObj(); } +} _error APT_UNUSED; + +#endif diff --git a/apt-pkg/contrib/extracttar.cc b/apt-pkg/contrib/extracttar.cc new file mode 100644 index 0000000..dc24375 --- /dev/null +++ b/apt-pkg/contrib/extracttar.cc @@ -0,0 +1,325 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + Extract a Tar - Tar Extractor + + Some performance measurements showed that zlib performed quite poorly + in comparison to a forked gzip process. This tar extractor makes use + of the fact that dup'd file descriptors have the same seek pointer + and that gzip will not read past the end of a compressed stream, + even if there is more data. We use the dup property to track extraction + progress and the gzip feature to just feed gzip a fd in the middle + of an AR file. + + ##################################################################### */ + /*}}}*/ +// Include Files /*{{{*/ +#include <config.h> + +#include <apt-pkg/configuration.h> +#include <apt-pkg/dirstream.h> +#include <apt-pkg/error.h> +#include <apt-pkg/extracttar.h> +#include <apt-pkg/fileutl.h> +#include <apt-pkg/strutl.h> + +#include <algorithm> +#include <iostream> +#include <string> +#include <fcntl.h> +#include <signal.h> +#include <string.h> +#include <unistd.h> + +#include <apti18n.h> + /*}}}*/ + +using namespace std; + +// The on disk header for a tar file. +struct ExtractTar::TarHeader +{ + char Name[100]; + char Mode[8]; + char UserID[8]; + char GroupID[8]; + char Size[12]; + char MTime[12]; + char Checksum[8]; + char LinkFlag; + char LinkName[100]; + char MagicNumber[8]; + char UserName[32]; + char GroupName[32]; + char Major[8]; + char Minor[8]; +}; + +// We need to read long names (names and link targets) into memory, so let's +// have a limit (shamelessly stolen from libarchive) to avoid people OOMing +// us with large streams. +static const unsigned long long APT_LONGNAME_LIMIT = 1048576llu; + +// A file size limit that we allow extracting. Currently, that's 128 GB. +// We also should leave some wiggle room for code adding files to it, and +// possibly conversion for signed, so this should not be larger than like 2**62. +static const unsigned long long APT_FILESIZE_LIMIT = 1llu << 37; + +// ExtractTar::ExtractTar - Constructor /*{{{*/ +// --------------------------------------------------------------------- +/* */ +ExtractTar::ExtractTar(FileFd &Fd,unsigned long long Max,string DecompressionProgram) + : File(Fd), MaxInSize(Max), DecompressProg(DecompressionProgram) +{ + GZPid = -1; + Eof = false; +} + /*}}}*/ +// ExtractTar::ExtractTar - Destructor /*{{{*/ +// --------------------------------------------------------------------- +/* */ +ExtractTar::~ExtractTar() +{ + // Error close + Done(); +} + /*}}}*/ +// ExtractTar::Done - Reap the gzip sub process /*{{{*/ +bool ExtractTar::Done() +{ + return InFd.Close(); +} + /*}}}*/ +// ExtractTar::StartGzip - Startup gzip /*{{{*/ +// --------------------------------------------------------------------- +/* This creates a gzip sub process that has its input as the file itself. + If this tar file is embedded into something like an ar file then + gzip will efficiently ignore the extra bits. */ +bool ExtractTar::StartGzip() +{ + if (DecompressProg.empty()) + { + InFd.OpenDescriptor(File.Fd(), FileFd::ReadOnly, FileFd::None, false); + return true; + } + + std::vector<APT::Configuration::Compressor> const compressors = APT::Configuration::getCompressors(); + std::vector<APT::Configuration::Compressor>::const_iterator compressor = compressors.begin(); + for (; compressor != compressors.end(); ++compressor) { + if (compressor->Name == DecompressProg) { + return InFd.OpenDescriptor(File.Fd(), FileFd::ReadOnly, *compressor, false); + } + } + + return _error->Error(_("Cannot find a configured compressor for '%s'"), + DecompressProg.c_str()); + +} + /*}}}*/ +// ExtractTar::Go - Perform extraction /*{{{*/ +// --------------------------------------------------------------------- +/* This reads each 512 byte block from the archive and extracts the header + information into the Item structure. Then it resolves the UID/GID and + invokes the correct processing function. */ +bool ExtractTar::Go(pkgDirStream &Stream) +{ + if (StartGzip() == false) + return false; + + // Loop over all blocks + string LastLongLink, ItemLink; + string LastLongName, ItemName; + while (1) + { + bool BadRecord = false; + unsigned char Block[512]; + if (InFd.Read(Block,sizeof(Block),true) == false) + return false; + + if (InFd.Eof() == true) + break; + + // Get the checksum + TarHeader *Tar = (TarHeader *)Block; + unsigned long CheckSum; + if (StrToNum(Tar->Checksum,CheckSum,sizeof(Tar->Checksum),8) == false) + return _error->Error(_("Corrupted archive")); + + /* Compute the checksum field. The actual checksum is blanked out + with spaces so it is not included in the computation */ + unsigned long NewSum = 0; + memset(Tar->Checksum,' ',sizeof(Tar->Checksum)); + for (int I = 0; I != sizeof(Block); I++) + NewSum += Block[I]; + + /* Check for a block of nulls - in this case we kill gzip, GNU tar + does this.. */ + if (NewSum == ' '*sizeof(Tar->Checksum)) + return Done(); + + if (NewSum != CheckSum) + return _error->Error(_("Tar checksum failed, archive corrupted")); + + // Decode all of the fields + pkgDirStream::Item Itm; + if (StrToNum(Tar->Mode,Itm.Mode,sizeof(Tar->Mode),8) == false || + (Base256ToNum(Tar->UserID,Itm.UID,8) == false && + StrToNum(Tar->UserID,Itm.UID,sizeof(Tar->UserID),8) == false) || + (Base256ToNum(Tar->GroupID,Itm.GID,8) == false && + StrToNum(Tar->GroupID,Itm.GID,sizeof(Tar->GroupID),8) == false) || + (Base256ToNum(Tar->Size,Itm.Size,12) == false && + StrToNum(Tar->Size,Itm.Size,sizeof(Tar->Size),8) == false) || + (Base256ToNum(Tar->MTime,Itm.MTime,12) == false && + StrToNum(Tar->MTime,Itm.MTime,sizeof(Tar->MTime),8) == false) || + StrToNum(Tar->Major,Itm.Major,sizeof(Tar->Major),8) == false || + StrToNum(Tar->Minor,Itm.Minor,sizeof(Tar->Minor),8) == false) + return _error->Error(_("Corrupted archive")); + + // Security check. Prevents overflows below the code when rounding up in skip/copy code, + // and provides modest protection against decompression bombs. + if (Itm.Size > APT_FILESIZE_LIMIT) + return _error->Error("Tar member too large: %llu > %llu bytes", Itm.Size, APT_FILESIZE_LIMIT); + + // Grab the filename and link target: use last long name if one was + // set, otherwise use the header value as-is, but remember that it may + // fill the entire 100-byte block and needs to be zero-terminated. + // See Debian Bug #689582. + if (LastLongName.empty() == false) + Itm.Name = (char *)LastLongName.c_str(); + else + Itm.Name = (char *)ItemName.assign(Tar->Name, sizeof(Tar->Name)).c_str(); + if (Itm.Name[0] == '.' && Itm.Name[1] == '/' && Itm.Name[2] != 0) + Itm.Name += 2; + + if (LastLongLink.empty() == false) + Itm.LinkTarget = (char *)LastLongLink.c_str(); + else + Itm.LinkTarget = (char *)ItemLink.assign(Tar->LinkName, sizeof(Tar->LinkName)).c_str(); + + // Convert the type over + switch (Tar->LinkFlag) + { + case NormalFile0: + case NormalFile: + Itm.Type = pkgDirStream::Item::File; + break; + + case HardLink: + Itm.Type = pkgDirStream::Item::HardLink; + break; + + case SymbolicLink: + Itm.Type = pkgDirStream::Item::SymbolicLink; + break; + + case CharacterDevice: + Itm.Type = pkgDirStream::Item::CharDevice; + break; + + case BlockDevice: + Itm.Type = pkgDirStream::Item::BlockDevice; + break; + + case Directory: + Itm.Type = pkgDirStream::Item::Directory; + break; + + case FIFO: + Itm.Type = pkgDirStream::Item::FIFO; + break; + + case GNU_LongLink: + { + unsigned long long Length = Itm.Size; + unsigned char Block[512]; + if (Length > APT_LONGNAME_LIMIT) + return _error->Error("Long name to large: %llu bytes > %llu bytes", Length, APT_LONGNAME_LIMIT); + while (Length > 0) + { + if (InFd.Read(Block,sizeof(Block),true) == false) + return false; + if (Length <= sizeof(Block)) + { + LastLongLink.append(Block,Block+sizeof(Block)); + break; + } + LastLongLink.append(Block,Block+sizeof(Block)); + Length -= sizeof(Block); + } + continue; + } + + case GNU_LongName: + { + unsigned long long Length = Itm.Size; + unsigned char Block[512]; + if (Length > APT_LONGNAME_LIMIT) + return _error->Error("Long name to large: %llu bytes > %llu bytes", Length, APT_LONGNAME_LIMIT); + while (Length > 0) + { + if (InFd.Read(Block,sizeof(Block),true) == false) + return false; + if (Length < sizeof(Block)) + { + LastLongName.append(Block,Block+sizeof(Block)); + break; + } + LastLongName.append(Block,Block+sizeof(Block)); + Length -= sizeof(Block); + } + continue; + } + + default: + BadRecord = true; + _error->Warning(_("Unknown TAR header type %u"), (unsigned)Tar->LinkFlag); + break; + } + + int Fd = -1; + if (not BadRecord && not Stream.DoItem(Itm, Fd)) + return false; + + if (Fd == -1 || Fd < -2 || BadRecord) + { + if (Itm.Size > 0 && not InFd.Skip(((Itm.Size + (sizeof(Block) - 1)) / sizeof(Block)) * sizeof(Block))) + return false; + } + else if (Itm.Size != 0) + { + // Copy the file over the FD + auto Size = Itm.Size; + unsigned char Junk[32*1024]; + do + { + auto const Read = std::min<unsigned long long>(Size, sizeof(Junk)); + if (not InFd.Read(Junk, ((Read + (sizeof(Block) - 1)) / sizeof(Block)) * sizeof(Block))) + return false; + + if (Fd > 0) + { + if (not FileFd::Write(Fd, Junk, Read)) + return Stream.Fail(Itm, Fd); + } + // An Fd of -2 means to send to a special processing function + else if (Fd == -2) + { + if (not Stream.Process(Itm, Junk, Read, Itm.Size - Size)) + return Stream.Fail(Itm, Fd); + } + + Size -= Read; + } while (Size != 0); + } + + // And finish up + if (not BadRecord && not Stream.FinishedFile(Itm, Fd)) + return false; + LastLongName.erase(); + LastLongLink.erase(); + } + + return Done(); +} + /*}}}*/ diff --git a/apt-pkg/contrib/extracttar.h b/apt-pkg/contrib/extracttar.h new file mode 100644 index 0000000..a3c862a --- /dev/null +++ b/apt-pkg/contrib/extracttar.h @@ -0,0 +1,55 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + Extract a Tar - Tar Extractor + + The tar extractor takes an ordinary gzip compressed tar stream from + the given file and explodes it, passing the individual items to the + given Directory Stream for processing. + + ##################################################################### */ + /*}}}*/ +#ifndef PKGLIB_EXTRACTTAR_H +#define PKGLIB_EXTRACTTAR_H + +#include <apt-pkg/fileutl.h> +#include <apt-pkg/macros.h> + +#include <string> + + +class pkgDirStream; + +class APT_PUBLIC ExtractTar +{ + protected: + + struct TarHeader; + + // The varios types items can be + enum ItemType {NormalFile0 = '\0',NormalFile = '0',HardLink = '1', + SymbolicLink = '2',CharacterDevice = '3', + BlockDevice = '4',Directory = '5',FIFO = '6', + GNU_LongLink = 'K',GNU_LongName = 'L'}; + + FileFd &File; + unsigned long long MaxInSize; + int GZPid; + FileFd InFd; + bool Eof; + std::string DecompressProg; + + // Fork and reap gzip + bool StartGzip(); + bool Done(); + + public: + + bool Go(pkgDirStream &Stream); + + ExtractTar(FileFd &Fd,unsigned long long Max,std::string DecompressionProgram); + virtual ~ExtractTar(); +}; + +#endif diff --git a/apt-pkg/contrib/fileutl.cc b/apt-pkg/contrib/fileutl.cc new file mode 100644 index 0000000..fad51de --- /dev/null +++ b/apt-pkg/contrib/fileutl.cc @@ -0,0 +1,3483 @@ +// -*- mode: cpp; mode: fold -*- +// SPDX-License-Identifier: GPL-2.0+ +// Description /*{{{*/ +/* ###################################################################### + + File Utilities + + CopyFile - Buffered copy of a single file + GetLock - dpkg compatible lock file manipulation (fcntl) + + This file had this historic note, but now includes further changes + under the GPL-2.0+: + + Most of this source is placed in the Public Domain, do with it what + you will + It was originally written by Jason Gunthorpe <jgg@debian.org>. + FileFd gzip support added by Martin Pitt <martin.pitt@canonical.com> + + The exception is RunScripts() it is under the GPLv2 + + We believe that this reference to GPLv2 was not meant to exclude later + versions as that would have changed the overall project license from GPL-2+ + to GPL-2. + + ##################################################################### */ + /*}}}*/ +// Include Files /*{{{*/ +#include <config.h> + +#include <apt-pkg/aptconfiguration.h> +#include <apt-pkg/configuration.h> +#include <apt-pkg/error.h> +#include <apt-pkg/fileutl.h> +#include <apt-pkg/macros.h> +#include <apt-pkg/pkgsystem.h> +#include <apt-pkg/strutl.h> + +#include <cstdio> +#include <cstdlib> +#include <cstring> +#include <iostream> +#include <string> +#include <vector> +#include <ctype.h> +#include <dirent.h> +#include <errno.h> +#include <fcntl.h> +#include <glob.h> +#include <grp.h> +#include <pwd.h> +#include <signal.h> +#include <stdarg.h> +#include <stddef.h> +#include <stdio.h> +#include <sys/select.h> +#include <sys/stat.h> +#include <sys/time.h> +#include <sys/wait.h> +#include <time.h> +#include <unistd.h> + +#include <algorithm> +#include <memory> +#include <set> + +#ifdef HAVE_ZLIB +#include <zlib.h> +#endif +#ifdef HAVE_BZ2 +#include <bzlib.h> +#endif +#ifdef HAVE_LZMA +#include <lzma.h> +#endif +#ifdef HAVE_LZ4 +#include <lz4frame.h> +#endif +#ifdef HAVE_ZSTD +#include <zstd.h> +#endif +#ifdef HAVE_SYSTEMD +#include <systemd/sd-bus.h> +#endif +#include <endian.h> +#include <stdint.h> + +#if __gnu_linux__ +#include <sys/prctl.h> +#endif + +#include <apti18n.h> + /*}}}*/ + +using namespace std; + +// RunScripts - Run a set of scripts from a configuration subtree /*{{{*/ +// --------------------------------------------------------------------- +/* */ +bool RunScripts(const char *Cnf) +{ + Configuration::Item const *Opts = _config->Tree(Cnf); + if (Opts == 0 || Opts->Child == 0) + return true; + Opts = Opts->Child; + + // Fork for running the system calls + pid_t Child = ExecFork(); + + // This is the child + if (Child == 0) + { + if (_system != nullptr && _system->IsLocked() == true && (stringcasecmp(Cnf, "dpkg::post-invoke") == 0 || stringcasecmp(Cnf, "dpkg::pre-invoke") == 0)) + setenv("DPKG_FRONTEND_LOCKED", "true", 1); + if (_config->FindDir("DPkg::Chroot-Directory","/") != "/") + { + std::cerr << "Chrooting into " + << _config->FindDir("DPkg::Chroot-Directory") + << std::endl; + if (chroot(_config->FindDir("DPkg::Chroot-Directory","/").c_str()) != 0) + _exit(100); + } + + if (chdir("/tmp/") != 0) + _exit(100); + + unsigned int Count = 1; + for (; Opts != 0; Opts = Opts->Next, Count++) + { + if (Opts->Value.empty() == true) + continue; + + if(_config->FindB("Debug::RunScripts", false) == true) + std::clog << "Running external script: '" + << Opts->Value << "'" << std::endl; + + if (system(Opts->Value.c_str()) != 0) + _exit(100+Count); + } + _exit(0); + } + + // Wait for the child + int Status = 0; + while (waitpid(Child,&Status,0) != Child) + { + if (errno == EINTR) + continue; + return _error->Errno("waitpid","Couldn't wait for subprocess"); + } + + // Check for an error code. + if (WIFEXITED(Status) == 0 || WEXITSTATUS(Status) != 0) + { + unsigned int Count = WEXITSTATUS(Status); + if (Count > 100) + { + Count -= 100; + for (; Opts != 0 && Count != 1; Opts = Opts->Next, Count--); + _error->Error("Problem executing scripts %s '%s'",Cnf,Opts->Value.c_str()); + } + + return _error->Error("Sub-process returned an error code"); + } + + return true; +} + /*}}}*/ + +// CopyFile - Buffered copy of a file /*{{{*/ +// --------------------------------------------------------------------- +/* The caller is expected to set things so that failure causes erasure */ +bool CopyFile(FileFd &From,FileFd &To) +{ + if (From.IsOpen() == false || To.IsOpen() == false || + From.Failed() == true || To.Failed() == true) + return false; + + // Buffered copy between fds + constexpr size_t BufSize = APT_BUFFER_SIZE; + std::unique_ptr<unsigned char[]> Buf(new unsigned char[BufSize]); + unsigned long long ToRead = 0; + do { + if (From.Read(Buf.get(),BufSize, &ToRead) == false || + To.Write(Buf.get(),ToRead) == false) + return false; + } while (ToRead != 0); + + return true; +} + /*}}}*/ +bool RemoveFileAt(char const * const Function, int const dirfd, std::string const &FileName)/*{{{*/ +{ + if (FileName == "/dev/null") + return true; + errno = 0; + if (unlinkat(dirfd, FileName.c_str(), 0) != 0) + { + if (errno == ENOENT) + return true; + + return _error->WarningE(Function,_("Problem unlinking the file %s"), FileName.c_str()); + } + return true; +} + /*}}}*/ +bool RemoveFile(char const * const Function, std::string const &FileName)/*{{{*/ +{ + if (FileName == "/dev/null") + return true; + errno = 0; + if (unlink(FileName.c_str()) != 0) + { + if (errno == ENOENT) + return true; + + return _error->WarningE(Function,_("Problem unlinking the file %s"), FileName.c_str()); + } + return true; +} + /*}}}*/ +// GetLock - Gets a lock file /*{{{*/ +// --------------------------------------------------------------------- +/* This will create an empty file of the given name and lock it. Once this + is done all other calls to GetLock in any other process will fail with + -1. The return result is the fd of the file, the call should call + close at some time. */ + +static std::string GetProcessName(int pid) +{ + struct HideError + { + int err; + HideError() : err(errno) { _error->PushToStack(); } + ~HideError() + { + errno = err; + _error->RevertToStack(); + } + } hideError; + std::string path; + strprintf(path, "/proc/%d/status", pid); + FileFd status(path, FileFd::ReadOnly); + std::string line; + while (status.ReadLine(line)) + { + if (line.substr(0, 5) == "Name:") + return line.substr(6); + } + return ""; +} +int GetLock(string File,bool Errors) +{ + // GetLock() is used in aptitude on directories with public-write access + // Use O_NOFOLLOW here to prevent symlink traversal attacks + int FD = open(File.c_str(),O_RDWR | O_CREAT | O_NOFOLLOW,0640); + if (FD < 0) + { + // Read only .. can't have locking problems there. + if (errno == EROFS) + { + _error->Warning(_("Not using locking for read only lock file %s"),File.c_str()); + return dup(0); // Need something for the caller to close + } + + if (Errors == true) + _error->Errno("open",_("Could not open lock file %s"),File.c_str()); + + // Feh.. We do this to distinguish the lock vs open case.. + errno = EPERM; + return -1; + } + SetCloseExec(FD,true); + + // Acquire a write lock + struct flock fl; + fl.l_type = F_WRLCK; + fl.l_whence = SEEK_SET; + fl.l_start = 0; + fl.l_len = 0; + if (fcntl(FD,F_SETLK,&fl) == -1) + { + // always close to not leak resources + int Tmp = errno; + + if ((errno == EACCES || errno == EAGAIN)) + { + fl.l_type = F_WRLCK; + fl.l_whence = SEEK_SET; + fl.l_start = 0; + fl.l_len = 0; + fl.l_pid = -1; + fcntl(FD, F_GETLK, &fl); + } + else + { + fl.l_pid = -1; + } + close(FD); + errno = Tmp; + + if (errno == ENOLCK) + { + _error->Warning(_("Not using locking for nfs mounted lock file %s"),File.c_str()); + return dup(0); // Need something for the caller to close + } + + if (Errors == true) + { + // We only do the lookup in the if ((errno == EACCES || errno == EAGAIN)) + // case, so we do not need to show the errno strerrr here... + if (fl.l_pid != -1) + { + auto name = GetProcessName(fl.l_pid); + if (name.empty()) + _error->Error(_("Could not get lock %s. It is held by process %d"), File.c_str(), fl.l_pid); + else + _error->Error(_("Could not get lock %s. It is held by process %d (%s)"), File.c_str(), fl.l_pid, name.c_str()); + } + else + _error->Errno("open", _("Could not get lock %s"), File.c_str()); + + _error->Notice(_("Be aware that removing the lock file is not a solution and may break your system.")); + } + + return -1; + } + + return FD; +} + /*}}}*/ +// FileExists - Check if a file exists /*{{{*/ +// --------------------------------------------------------------------- +/* Beware: Directories are also files! */ +bool FileExists(string File) +{ + struct stat Buf; + if (stat(File.c_str(),&Buf) != 0) + return false; + return true; +} + /*}}}*/ +// RealFileExists - Check if a file exists and if it is really a file /*{{{*/ +// --------------------------------------------------------------------- +/* */ +bool RealFileExists(string File) +{ + struct stat Buf; + if (stat(File.c_str(),&Buf) != 0) + return false; + return ((Buf.st_mode & S_IFREG) != 0); +} + /*}}}*/ +// DirectoryExists - Check if a directory exists and is really one /*{{{*/ +// --------------------------------------------------------------------- +/* */ +bool DirectoryExists(string const &Path) +{ + struct stat Buf; + if (stat(Path.c_str(),&Buf) != 0) + return false; + return ((Buf.st_mode & S_IFDIR) != 0); +} + /*}}}*/ +// CreateDirectory - poor man's mkdir -p guarded by a parent directory /*{{{*/ +// --------------------------------------------------------------------- +/* This method will create all directories needed for path in good old + mkdir -p style but refuses to do this if Parent is not a prefix of + this Path. Example: /var/cache/ and /var/cache/apt/archives are given, + so it will create apt/archives if /var/cache exists - on the other + hand if the parent is /var/lib the creation will fail as this path + is not a parent of the path to be generated. */ +bool CreateDirectory(string const &Parent, string const &Path) +{ + if (Parent.empty() == true || Path.empty() == true) + return false; + + if (DirectoryExists(Path) == true) + return true; + + if (DirectoryExists(Parent) == false) + return false; + + // we are not going to create directories "into the blue" + if (Path.compare(0, Parent.length(), Parent) != 0) + return false; + + vector<string> const dirs = VectorizeString(Path.substr(Parent.size()), '/'); + string progress = Parent; + for (vector<string>::const_iterator d = dirs.begin(); d != dirs.end(); ++d) + { + if (d->empty() == true) + continue; + + progress.append("/").append(*d); + if (DirectoryExists(progress) == true) + continue; + + if (mkdir(progress.c_str(), 0755) != 0) + return false; + } + return true; +} + /*}}}*/ +// CreateAPTDirectoryIfNeeded - ensure that the given directory exists /*{{{*/ +// --------------------------------------------------------------------- +/* a small wrapper around CreateDirectory to check if it exists and to + remove the trailing "/apt/" from the parent directory if needed */ +bool CreateAPTDirectoryIfNeeded(string const &Parent, string const &Path) +{ + if (DirectoryExists(Path) == true) + return true; + + size_t const len = Parent.size(); + if (len > 5 && Parent.find("/apt/", len - 6, 5) == len - 5) + { + if (CreateDirectory(Parent.substr(0,len-5), Path) == true) + return true; + } + else if (CreateDirectory(Parent, Path) == true) + return true; + + return false; +} + /*}}}*/ +// GetListOfFilesInDir - returns a vector of files in the given dir /*{{{*/ +// --------------------------------------------------------------------- +/* If an extension is given only files with this extension are included + in the returned vector, otherwise every "normal" file is included. */ +std::vector<string> GetListOfFilesInDir(string const &Dir, string const &Ext, + bool const &SortList, bool const &AllowNoExt) +{ + std::vector<string> ext; + ext.reserve(2); + if (Ext.empty() == false) + ext.push_back(Ext); + if (AllowNoExt == true && ext.empty() == false) + ext.push_back(""); + return GetListOfFilesInDir(Dir, ext, SortList); +} +std::vector<string> GetListOfFilesInDir(string const &Dir, std::vector<string> const &Ext, + bool const &SortList) +{ + // Attention debuggers: need to be set with the environment config file! + bool const Debug = _config->FindB("Debug::GetListOfFilesInDir", false); + if (Debug == true) + { + std::clog << "Accept in " << Dir << " only files with the following " << Ext.size() << " extensions:" << std::endl; + if (Ext.empty() == true) + std::clog << "\tNO extension" << std::endl; + else + for (std::vector<string>::const_iterator e = Ext.begin(); + e != Ext.end(); ++e) + std::clog << '\t' << (e->empty() == true ? "NO" : *e) << " extension" << std::endl; + } + + std::vector<string> List; + + if (DirectoryExists(Dir) == false) + { + _error->Error(_("List of files can't be created as '%s' is not a directory"), Dir.c_str()); + return List; + } + + Configuration::MatchAgainstConfig SilentIgnore("Dir::Ignore-Files-Silently"); + DIR *D = opendir(Dir.c_str()); + if (D == 0) + { + if (errno == EACCES) + _error->WarningE("opendir", _("Unable to read %s"), Dir.c_str()); + else + _error->Errno("opendir", _("Unable to read %s"), Dir.c_str()); + return List; + } + + for (struct dirent *Ent = readdir(D); Ent != 0; Ent = readdir(D)) + { + // skip "hidden" files + if (Ent->d_name[0] == '.') + continue; + + // Make sure it is a file and not something else + string const File = flCombine(Dir,Ent->d_name); +#ifdef _DIRENT_HAVE_D_TYPE + if (Ent->d_type != DT_REG) +#endif + { + if (RealFileExists(File) == false) + { + // do not show ignoration warnings for directories + if ( +#ifdef _DIRENT_HAVE_D_TYPE + Ent->d_type == DT_DIR || +#endif + DirectoryExists(File) == true) + continue; + if (SilentIgnore.Match(Ent->d_name) == false) + _error->Notice(_("Ignoring '%s' in directory '%s' as it is not a regular file"), Ent->d_name, Dir.c_str()); + continue; + } + } + + // check for accepted extension: + // no extension given -> periods are bad as hell! + // extensions given -> "" extension allows no extension + if (Ext.empty() == false) + { + string d_ext = flExtension(Ent->d_name); + if (d_ext == Ent->d_name) // no extension + { + if (std::find(Ext.begin(), Ext.end(), "") == Ext.end()) + { + if (Debug == true) + std::clog << "Bad file: " << Ent->d_name << " → no extension" << std::endl; + if (SilentIgnore.Match(Ent->d_name) == false) + _error->Notice(_("Ignoring file '%s' in directory '%s' as it has no filename extension"), Ent->d_name, Dir.c_str()); + continue; + } + } + else if (std::find(Ext.begin(), Ext.end(), d_ext) == Ext.end()) + { + if (Debug == true) + std::clog << "Bad file: " << Ent->d_name << " → bad extension »" << flExtension(Ent->d_name) << "«" << std::endl; + if (SilentIgnore.Match(Ent->d_name) == false) + _error->Notice(_("Ignoring file '%s' in directory '%s' as it has an invalid filename extension"), Ent->d_name, Dir.c_str()); + continue; + } + } + + // Skip bad filenames ala run-parts + const char *C = Ent->d_name; + for (; *C != 0; ++C) + if (isalpha(*C) == 0 && isdigit(*C) == 0 + && *C != '_' && *C != '-' && *C != ':') { + // no required extension -> dot is a bad character + if (*C == '.' && Ext.empty() == false) + continue; + break; + } + + // we don't reach the end of the name -> bad character included + if (*C != 0) + { + if (Debug == true) + std::clog << "Bad file: " << Ent->d_name << " → bad character »" + << *C << "« in filename (period allowed: " << (Ext.empty() ? "no" : "yes") << ")" << std::endl; + continue; + } + + // skip filenames which end with a period. These are never valid + if (*(C - 1) == '.') + { + if (Debug == true) + std::clog << "Bad file: " << Ent->d_name << " → Period as last character" << std::endl; + continue; + } + + if (Debug == true) + std::clog << "Accept file: " << Ent->d_name << " in " << Dir << std::endl; + List.push_back(File); + } + closedir(D); + + if (SortList == true) + std::sort(List.begin(),List.end()); + return List; +} +std::vector<string> GetListOfFilesInDir(string const &Dir, bool SortList) +{ + bool const Debug = _config->FindB("Debug::GetListOfFilesInDir", false); + if (Debug == true) + std::clog << "Accept in " << Dir << " all regular files" << std::endl; + + std::vector<string> List; + + if (DirectoryExists(Dir) == false) + { + _error->Error(_("List of files can't be created as '%s' is not a directory"), Dir.c_str()); + return List; + } + + DIR *D = opendir(Dir.c_str()); + if (D == 0) + { + _error->Errno("opendir",_("Unable to read %s"),Dir.c_str()); + return List; + } + + for (struct dirent *Ent = readdir(D); Ent != 0; Ent = readdir(D)) + { + // skip "hidden" files + if (Ent->d_name[0] == '.') + continue; + + // Make sure it is a file and not something else + string const File = flCombine(Dir,Ent->d_name); +#ifdef _DIRENT_HAVE_D_TYPE + if (Ent->d_type != DT_REG) +#endif + { + if (RealFileExists(File) == false) + { + if (Debug == true) + std::clog << "Bad file: " << Ent->d_name << " → it is not a real file" << std::endl; + continue; + } + } + + // Skip bad filenames ala run-parts + const char *C = Ent->d_name; + for (; *C != 0; ++C) + if (isalpha(*C) == 0 && isdigit(*C) == 0 + && *C != '_' && *C != '-' && *C != '.') + break; + + // we don't reach the end of the name -> bad character included + if (*C != 0) + { + if (Debug == true) + std::clog << "Bad file: " << Ent->d_name << " → bad character »" << *C << "« in filename" << std::endl; + continue; + } + + // skip filenames which end with a period. These are never valid + if (*(C - 1) == '.') + { + if (Debug == true) + std::clog << "Bad file: " << Ent->d_name << " → Period as last character" << std::endl; + continue; + } + + if (Debug == true) + std::clog << "Accept file: " << Ent->d_name << " in " << Dir << std::endl; + List.push_back(File); + } + closedir(D); + + if (SortList == true) + std::sort(List.begin(),List.end()); + return List; +} + /*}}}*/ +// SafeGetCWD - This is a safer getcwd that returns a dynamic string /*{{{*/ +// --------------------------------------------------------------------- +/* We return / on failure. */ +string SafeGetCWD() +{ + // Stash the current dir. + char S[300]; + S[0] = 0; + if (getcwd(S,sizeof(S)-2) == 0) + return "/"; + unsigned int Len = strlen(S); + S[Len] = '/'; + S[Len+1] = 0; + return S; +} + /*}}}*/ +// GetModificationTime - Get the mtime of the given file or -1 on error /*{{{*/ +// --------------------------------------------------------------------- +/* We return / on failure. */ +time_t GetModificationTime(string const &Path) +{ + struct stat St; + if (stat(Path.c_str(), &St) < 0) + return -1; + return St.st_mtime; +} + /*}}}*/ +// flNotDir - Strip the directory from the filename /*{{{*/ +// --------------------------------------------------------------------- +/* */ +string flNotDir(string File) +{ + string::size_type Res = File.rfind('/'); + if (Res == string::npos) + return File; + Res++; + return string(File,Res,Res - File.length()); +} + /*}}}*/ +// flNotFile - Strip the file from the directory name /*{{{*/ +// --------------------------------------------------------------------- +/* Result ends in a / */ +string flNotFile(string File) +{ + string::size_type Res = File.rfind('/'); + if (Res == string::npos) + return "./"; + Res++; + return string(File,0,Res); +} + /*}}}*/ +// flExtension - Return the extension for the file /*{{{*/ +// --------------------------------------------------------------------- +/* */ +string flExtension(string File) +{ + string::size_type Res = File.rfind('.'); + if (Res == string::npos) + return File; + Res++; + return string(File,Res); +} + /*}}}*/ +// flNoLink - If file is a symlink then deref it /*{{{*/ +// --------------------------------------------------------------------- +/* If the name is not a link then the returned path is the input. */ +string flNoLink(string File) +{ + struct stat St; + if (lstat(File.c_str(),&St) != 0 || S_ISLNK(St.st_mode) == 0) + return File; + if (stat(File.c_str(),&St) != 0) + return File; + + /* Loop resolving the link. There is no need to limit the number of + loops because the stat call above ensures that the symlink is not + circular */ + char Buffer[1024]; + string NFile = File; + while (1) + { + // Read the link + ssize_t Res; + if ((Res = readlink(NFile.c_str(),Buffer,sizeof(Buffer))) <= 0 || + (size_t)Res >= sizeof(Buffer)) + return File; + + // Append or replace the previous path + Buffer[Res] = 0; + if (Buffer[0] == '/') + NFile = Buffer; + else + NFile = flNotFile(NFile) + Buffer; + + // See if we are done + if (lstat(NFile.c_str(),&St) != 0) + return File; + if (S_ISLNK(St.st_mode) == 0) + return NFile; + } +} + /*}}}*/ +// flCombine - Combine a file and a directory /*{{{*/ +// --------------------------------------------------------------------- +/* If the file is an absolute path then it is just returned, otherwise + the directory is pre-pended to it. */ +string flCombine(string Dir,string File) +{ + if (File.empty() == true) + return string(); + + if (File[0] == '/' || Dir.empty() == true) + return File; + if (File.length() >= 2 && File[0] == '.' && File[1] == '/') + return File; + if (Dir[Dir.length()-1] == '/') + return Dir + File; + return Dir + '/' + File; +} + /*}}}*/ +// flAbsPath - Return the absolute path of the filename /*{{{*/ +// --------------------------------------------------------------------- +/* */ +string flAbsPath(string File) +{ + char *p = realpath(File.c_str(), NULL); + if (p == NULL) + { + _error->Errno("realpath", "flAbsPath on %s failed", File.c_str()); + return ""; + } + std::string AbsPath(p); + free(p); + return AbsPath; +} + /*}}}*/ +std::string flNormalize(std::string file) /*{{{*/ +{ + if (file.empty()) + return file; + // do some normalisation by removing // and /./ from the path + size_t found = string::npos; + while ((found = file.find("/./")) != string::npos) + file.replace(found, 3, "/"); + while ((found = file.find("//")) != string::npos) + file.replace(found, 2, "/"); + + if (APT::String::Startswith(file, "/dev/null")) + { + file.erase(strlen("/dev/null")); + return file; + } + return file; +} + /*}}}*/ +// SetCloseExec - Set the close on exec flag /*{{{*/ +// --------------------------------------------------------------------- +/* */ +void SetCloseExec(int Fd,bool Close) +{ + if (fcntl(Fd,F_SETFD,(Close == false)?0:FD_CLOEXEC) != 0) + { + cerr << "FATAL -> Could not set close on exec " << strerror(errno) << endl; + exit(100); + } +} + /*}}}*/ +// SetNonBlock - Set the nonblocking flag /*{{{*/ +// --------------------------------------------------------------------- +/* */ +void SetNonBlock(int Fd,bool Block) +{ + int Flags = fcntl(Fd,F_GETFL) & (~O_NONBLOCK); + if (fcntl(Fd,F_SETFL,Flags | ((Block == false)?0:O_NONBLOCK)) != 0) + { + cerr << "FATAL -> Could not set non-blocking flag " << strerror(errno) << endl; + exit(100); + } +} + /*}}}*/ +// WaitFd - Wait for a FD to become readable /*{{{*/ +// --------------------------------------------------------------------- +/* This waits for a FD to become readable using select. It is useful for + applications making use of non-blocking sockets. The timeout is + in seconds. */ +bool WaitFd(int Fd,bool write,unsigned long timeout) +{ + fd_set Set; + struct timeval tv; + FD_ZERO(&Set); + FD_SET(Fd,&Set); + tv.tv_sec = timeout; + tv.tv_usec = 0; + if (write == true) + { + int Res; + do + { + Res = select(Fd+1,0,&Set,0,(timeout != 0?&tv:0)); + } + while (Res < 0 && errno == EINTR); + + if (Res <= 0) + return false; + } + else + { + int Res; + do + { + Res = select(Fd+1,&Set,0,0,(timeout != 0?&tv:0)); + } + while (Res < 0 && errno == EINTR); + + if (Res <= 0) + return false; + } + + return true; +} + /*}}}*/ +// MergeKeepFdsFromConfiguration - Merge APT::Keep-Fds configuration /*{{{*/ +// --------------------------------------------------------------------- +/* This is used to merge the APT::Keep-Fds with the provided KeepFDs + * set. + */ +void MergeKeepFdsFromConfiguration(std::set<int> &KeepFDs) +{ + Configuration::Item const *Opts = _config->Tree("APT::Keep-Fds"); + if (Opts != 0 && Opts->Child != 0) + { + Opts = Opts->Child; + for (; Opts != 0; Opts = Opts->Next) + { + if (Opts->Value.empty() == true) + continue; + int fd = atoi(Opts->Value.c_str()); + KeepFDs.insert(fd); + } + } +} + /*}}}*/ +// ExecFork - Magical fork that sanitizes the context before execing /*{{{*/ +// --------------------------------------------------------------------- +/* This is used if you want to cleanse the environment for the forked + child, it fixes up the important signals and nukes all of the fds, + otherwise acts like normal fork. */ +pid_t ExecFork() +{ + set<int> KeepFDs; + // we need to merge the Keep-Fds as external tools like + // debconf-apt-progress use it + MergeKeepFdsFromConfiguration(KeepFDs); + return ExecFork(KeepFDs); +} + +pid_t ExecFork(std::set<int> KeepFDs) +{ + // Fork off the process + pid_t Process = fork(); + if (Process < 0) + { + cerr << "FATAL -> Failed to fork." << endl; + exit(100); + } + + // Spawn the subprocess + if (Process == 0) + { + // Setup the signals + signal(SIGPIPE,SIG_DFL); + signal(SIGQUIT,SIG_DFL); + signal(SIGINT,SIG_DFL); + signal(SIGWINCH,SIG_DFL); + signal(SIGCONT,SIG_DFL); + signal(SIGTSTP,SIG_DFL); + + DIR *dir = opendir("/proc/self/fd"); + if (dir != NULL) + { + struct dirent *ent; + while ((ent = readdir(dir))) + { + int fd = atoi(ent->d_name); + // If fd > 0, it was a fd number and not . or .. + if (fd >= 3 && KeepFDs.find(fd) == KeepFDs.end()) + fcntl(fd,F_SETFD,FD_CLOEXEC); + } + closedir(dir); + } else { + long ScOpenMax = sysconf(_SC_OPEN_MAX); + // Close all of our FDs - just in case + for (int K = 3; K != ScOpenMax; K++) + { + if(KeepFDs.find(K) == KeepFDs.end()) + fcntl(K,F_SETFD,FD_CLOEXEC); + } + } + } + + return Process; +} + /*}}}*/ +// ExecWait - Fancy waitpid /*{{{*/ +// --------------------------------------------------------------------- +/* Waits for the given sub process. If Reap is set then no errors are + generated. Otherwise a failed subprocess will generate a proper descriptive + message */ +bool ExecWait(pid_t Pid,const char *Name,bool Reap) +{ + if (Pid <= 1) + return true; + + // Wait and collect the error code + int Status; + while (waitpid(Pid,&Status,0) != Pid) + { + if (errno == EINTR) + continue; + + if (Reap == true) + return false; + + return _error->Error(_("Waited for %s but it wasn't there"),Name); + } + + + // Check for an error code. + if (WIFEXITED(Status) == 0 || WEXITSTATUS(Status) != 0) + { + if (Reap == true) + return false; + if (WIFSIGNALED(Status) != 0) + { + if( WTERMSIG(Status) == SIGSEGV) + return _error->Error(_("Sub-process %s received a segmentation fault."),Name); + else + return _error->Error(_("Sub-process %s received signal %u."),Name, WTERMSIG(Status)); + } + + if (WIFEXITED(Status) != 0) + return _error->Error(_("Sub-process %s returned an error code (%u)"),Name,WEXITSTATUS(Status)); + + return _error->Error(_("Sub-process %s exited unexpectedly"),Name); + } + + return true; +} + /*}}}*/ +// StartsWithGPGClearTextSignature - Check if a file is Pgp/GPG clearsigned /*{{{*/ +bool StartsWithGPGClearTextSignature(string const &FileName) +{ + FILE* gpg = fopen(FileName.c_str(), "r"); + if (gpg == nullptr) + return false; + + char * lineptr = nullptr; + size_t n = 0; + errno = 0; + ssize_t const result = getline(&lineptr, &n, gpg); + if (errno != 0) + { + _error->Errno("getline", "Could not read from %s", FileName.c_str()); + fclose(gpg); + free(lineptr); + return false; + } + fclose(gpg); + + _strrstrip(lineptr); + static const char* SIGMSG = "-----BEGIN PGP SIGNED MESSAGE-----"; + if (result == -1 || strcmp(lineptr, SIGMSG) != 0) + { + free(lineptr); + return false; + } + free(lineptr); + return true; +} + /*}}}*/ +// ChangeOwnerAndPermissionOfFile - set file attributes to requested values /*{{{*/ +bool ChangeOwnerAndPermissionOfFile(char const * const requester, char const * const file, char const * const user, char const * const group, mode_t const mode) +{ + if (strcmp(file, "/dev/null") == 0) + return true; + bool Res = true; + if (getuid() == 0 && strlen(user) != 0 && strlen(group) != 0) // if we aren't root, we can't chown, so don't try it + { + // ensure the file is owned by root and has good permissions + struct passwd const * const pw = getpwnam(user); + struct group const * const gr = getgrnam(group); + if (pw != NULL && gr != NULL && lchown(file, pw->pw_uid, gr->gr_gid) != 0) + Res &= _error->WarningE(requester, "chown to %s:%s of file %s failed", user, group, file); + } + struct stat Buf; + if (lstat(file, &Buf) != 0 || S_ISLNK(Buf.st_mode)) + return Res; + if (chmod(file, mode) != 0) + Res &= _error->WarningE(requester, "chmod 0%o of file %s failed", mode, file); + return Res; +} + /*}}}*/ + +struct APT_HIDDEN simple_buffer { /*{{{*/ + size_t buffersize_max = 0; + unsigned long long bufferstart = 0; + unsigned long long bufferend = 0; + char *buffer = nullptr; + + simple_buffer() { + reset(4096); + } + ~simple_buffer() { + delete[] buffer; + } + + const char *get() const { return buffer + bufferstart; } + char *get() { return buffer + bufferstart; } + const char *getend() const { return buffer + bufferend; } + char *getend() { return buffer + bufferend; } + bool empty() const { return bufferend <= bufferstart; } + bool full() const { return bufferend == buffersize_max; } + unsigned long long free() const { return buffersize_max - bufferend; } + unsigned long long size() const { return bufferend-bufferstart; } + void reset(size_t size) + { + if (size > buffersize_max) { + delete[] buffer; + buffersize_max = size; + buffer = new char[size]; + } + reset(); + } + void reset() { bufferend = bufferstart = 0; } + ssize_t read(void *to, unsigned long long requested_size) APT_MUSTCHECK + { + if (size() < requested_size) + requested_size = size(); + memcpy(to, buffer + bufferstart, requested_size); + bufferstart += requested_size; + if (bufferstart == bufferend) + bufferstart = bufferend = 0; + return requested_size; + } + ssize_t write(const void *from, unsigned long long requested_size) APT_MUSTCHECK + { + if (free() < requested_size) + requested_size = free(); + memcpy(getend(), from, requested_size); + bufferend += requested_size; + if (bufferstart == bufferend) + bufferstart = bufferend = 0; + return requested_size; + } +}; + /*}}}*/ + +class APT_HIDDEN FileFdPrivate { /*{{{*/ + friend class BufferedWriteFileFdPrivate; +protected: + FileFd * const filefd; + simple_buffer buffer; + int compressed_fd; + pid_t compressor_pid; + bool is_pipe; + APT::Configuration::Compressor compressor; + unsigned int openmode; + unsigned long long seekpos; +public: + + explicit FileFdPrivate(FileFd * const pfilefd) : filefd(pfilefd), + compressed_fd(-1), compressor_pid(-1), is_pipe(false), + openmode(0), seekpos(0) {}; + virtual APT::Configuration::Compressor get_compressor() const + { + return compressor; + } + virtual void set_compressor(APT::Configuration::Compressor const &compressor) + { + this->compressor = compressor; + } + virtual unsigned int get_openmode() const + { + return openmode; + } + virtual void set_openmode(unsigned int openmode) + { + this->openmode = openmode; + } + virtual bool get_is_pipe() const + { + return is_pipe; + } + virtual void set_is_pipe(bool is_pipe) + { + this->is_pipe = is_pipe; + } + virtual unsigned long long get_seekpos() const + { + return seekpos; + } + virtual void set_seekpos(unsigned long long seekpos) + { + this->seekpos = seekpos; + } + + virtual bool InternalOpen(int const iFd, unsigned int const Mode) = 0; + ssize_t InternalRead(void * To, unsigned long long Size) + { + // Drain the buffer if needed. + if (buffer.empty() == false) + { + return buffer.read(To, Size); + } + return InternalUnbufferedRead(To, Size); + } + virtual ssize_t InternalUnbufferedRead(void * const To, unsigned long long const Size) = 0; + virtual bool InternalReadError() { return filefd->FileFdErrno("read",_("Read error")); } + virtual char * InternalReadLine(char * To, unsigned long long Size) + { + if (unlikely(Size == 0)) + return nullptr; + // Read one byte less than buffer size to have space for trailing 0. + --Size; + + char * const InitialTo = To; + + while (Size > 0) { + if (buffer.empty() == true) + { + buffer.reset(); + unsigned long long actualread = 0; + if (filefd->Read(buffer.getend(), buffer.free(), &actualread) == false) + return nullptr; + buffer.bufferend = actualread; + if (buffer.size() == 0) + { + if (To == InitialTo) + return nullptr; + break; + } + filefd->Flags &= ~FileFd::HitEof; + } + + unsigned long long const OutputSize = std::min(Size, buffer.size()); + char const * const newline = static_cast<char const *>(memchr(buffer.get(), '\n', OutputSize)); + // Read until end of line or up to Size bytes from the buffer. + unsigned long long actualread = buffer.read(To, + (newline != nullptr) + ? (newline - buffer.get()) + 1 + : OutputSize); + To += actualread; + Size -= actualread; + if (newline != nullptr) + break; + } + *To = '\0'; + return InitialTo; + } + virtual bool InternalFlush() + { + return true; + } + virtual ssize_t InternalWrite(void const * const From, unsigned long long const Size) = 0; + virtual bool InternalWriteError() { return filefd->FileFdErrno("write",_("Write error")); } + virtual bool InternalSeek(unsigned long long const To) + { + // Our poor man seeking is costly, so try to avoid it + unsigned long long const iseekpos = filefd->Tell(); + if (iseekpos == To) + return true; + else if (iseekpos < To) + return filefd->Skip(To - iseekpos); + + if ((openmode & FileFd::ReadOnly) != FileFd::ReadOnly) + return filefd->FileFdError("Reopen is only implemented for read-only files!"); + InternalClose(filefd->FileName); + if (filefd->iFd != -1) + close(filefd->iFd); + filefd->iFd = -1; + if (filefd->TemporaryFileName.empty() == false) + filefd->iFd = open(filefd->TemporaryFileName.c_str(), O_RDONLY); + else if (filefd->FileName.empty() == false) + filefd->iFd = open(filefd->FileName.c_str(), O_RDONLY); + else + { + if (compressed_fd > 0) + if (lseek(compressed_fd, 0, SEEK_SET) != 0) + filefd->iFd = compressed_fd; + if (filefd->iFd < 0) + return filefd->FileFdError("Reopen is not implemented for pipes opened with FileFd::OpenDescriptor()!"); + } + + if (filefd->OpenInternDescriptor(openmode, compressor) == false) + return filefd->FileFdError("Seek on file %s because it couldn't be reopened", filefd->FileName.c_str()); + + buffer.reset(); + set_seekpos(0); + if (To != 0) + return filefd->Skip(To); + + seekpos = To; + return true; + } + virtual bool InternalSkip(unsigned long long Over) + { + unsigned long long constexpr buffersize = 1024; + char buffer[buffersize]; + while (Over != 0) + { + unsigned long long toread = std::min(buffersize, Over); + if (filefd->Read(buffer, toread) == false) + return filefd->FileFdError("Unable to seek ahead %llu",Over); + Over -= toread; + } + return true; + } + virtual bool InternalTruncate(unsigned long long const) + { + return filefd->FileFdError("Truncating compressed files is not implemented (%s)", filefd->FileName.c_str()); + } + virtual unsigned long long InternalTell() + { + // In theory, we could just return seekpos here always instead of + // seeking around, but not all users of FileFd use always Seek() and co + // so d->seekpos isn't always true and we can just use it as a hint if + // we have nothing else, but not always as an authority… + return seekpos - buffer.size(); + } + virtual unsigned long long InternalSize() + { + unsigned long long size = 0; + unsigned long long const oldSeek = filefd->Tell(); + unsigned long long constexpr ignoresize = 1024; + char ignore[ignoresize]; + unsigned long long read = 0; + do { + if (filefd->Read(ignore, ignoresize, &read) == false) + { + filefd->Seek(oldSeek); + return 0; + } + } while(read != 0); + size = filefd->Tell(); + filefd->Seek(oldSeek); + return size; + } + virtual bool InternalClose(std::string const &FileName) = 0; + virtual bool InternalStream() const { return false; } + virtual bool InternalAlwaysAutoClose() const { return true; } + + virtual ~FileFdPrivate() {} +}; + /*}}}*/ +class APT_HIDDEN BufferedWriteFileFdPrivate : public FileFdPrivate { /*{{{*/ +protected: + FileFdPrivate *wrapped; + simple_buffer writebuffer; + +public: + + explicit BufferedWriteFileFdPrivate(FileFdPrivate *Priv) : + FileFdPrivate(Priv->filefd), wrapped(Priv) {}; + + virtual APT::Configuration::Compressor get_compressor() const APT_OVERRIDE + { + return wrapped->get_compressor(); + } + virtual void set_compressor(APT::Configuration::Compressor const &compressor) APT_OVERRIDE + { + return wrapped->set_compressor(compressor); + } + virtual unsigned int get_openmode() const APT_OVERRIDE + { + return wrapped->get_openmode(); + } + virtual void set_openmode(unsigned int openmode) APT_OVERRIDE + { + return wrapped->set_openmode(openmode); + } + virtual bool get_is_pipe() const APT_OVERRIDE + { + return wrapped->get_is_pipe(); + } + virtual void set_is_pipe(bool is_pipe) APT_OVERRIDE + { + FileFdPrivate::set_is_pipe(is_pipe); + wrapped->set_is_pipe(is_pipe); + } + virtual unsigned long long get_seekpos() const APT_OVERRIDE + { + return wrapped->get_seekpos(); + } + virtual void set_seekpos(unsigned long long seekpos) APT_OVERRIDE + { + return wrapped->set_seekpos(seekpos); + } + virtual bool InternalOpen(int const iFd, unsigned int const Mode) APT_OVERRIDE + { + if (InternalFlush() == false) + return false; + return wrapped->InternalOpen(iFd, Mode); + } + virtual ssize_t InternalUnbufferedRead(void * const To, unsigned long long const Size) APT_OVERRIDE + { + if (InternalFlush() == false) + return -1; + return wrapped->InternalUnbufferedRead(To, Size); + + } + virtual bool InternalReadError() APT_OVERRIDE + { + return wrapped->InternalReadError(); + } + virtual char * InternalReadLine(char * To, unsigned long long Size) APT_OVERRIDE + { + if (InternalFlush() == false) + return nullptr; + return wrapped->InternalReadLine(To, Size); + } + virtual bool InternalFlush() APT_OVERRIDE + { + while (writebuffer.empty() == false) { + auto written = wrapped->InternalWrite(writebuffer.get(), + writebuffer.size()); + // Ignore interrupted syscalls + if (written < 0 && errno == EINTR) + continue; + if (written < 0) + return wrapped->InternalWriteError(); + + writebuffer.bufferstart += written; + } + writebuffer.reset(); + return wrapped->InternalFlush(); + } + virtual ssize_t InternalWrite(void const * const From, unsigned long long const Size) APT_OVERRIDE + { + // Optimisation: If the buffer is empty and we have more to write than + // would fit in the buffer (or equal number of bytes), write directly. + if (writebuffer.empty() == true && Size >= writebuffer.free()) + return wrapped->InternalWrite(From, Size); + + // Write as much into the buffer as possible and then flush if needed + auto written = writebuffer.write(From, Size); + + if (writebuffer.full() && InternalFlush() == false) + return -1; + + return written; + } + virtual bool InternalWriteError() APT_OVERRIDE + { + return wrapped->InternalWriteError(); + } + virtual bool InternalSeek(unsigned long long const To) APT_OVERRIDE + { + if (InternalFlush() == false) + return false; + return wrapped->InternalSeek(To); + } + virtual bool InternalSkip(unsigned long long Over) APT_OVERRIDE + { + if (InternalFlush() == false) + return false; + return wrapped->InternalSkip(Over); + } + virtual bool InternalTruncate(unsigned long long const Size) APT_OVERRIDE + { + if (InternalFlush() == false) + return false; + return wrapped->InternalTruncate(Size); + } + virtual unsigned long long InternalTell() APT_OVERRIDE + { + if (InternalFlush() == false) + return -1; + return wrapped->InternalTell(); + } + virtual unsigned long long InternalSize() APT_OVERRIDE + { + if (InternalFlush() == false) + return -1; + return wrapped->InternalSize(); + } + virtual bool InternalClose(std::string const &FileName) APT_OVERRIDE + { + return wrapped->InternalClose(FileName); + } + virtual bool InternalAlwaysAutoClose() const APT_OVERRIDE + { + return wrapped->InternalAlwaysAutoClose(); + } + virtual ~BufferedWriteFileFdPrivate() + { + delete wrapped; + } +}; + /*}}}*/ +class APT_HIDDEN GzipFileFdPrivate: public FileFdPrivate { /*{{{*/ +#ifdef HAVE_ZLIB +public: + gzFile gz; + virtual bool InternalOpen(int const iFd, unsigned int const Mode) APT_OVERRIDE + { + if ((Mode & FileFd::ReadWrite) == FileFd::ReadWrite) + gz = gzdopen(iFd, "r+"); + else if ((Mode & FileFd::WriteOnly) == FileFd::WriteOnly) + gz = gzdopen(iFd, "w"); + else + gz = gzdopen(iFd, "r"); + filefd->Flags |= FileFd::Compressed; + return gz != nullptr; + } + virtual ssize_t InternalUnbufferedRead(void * const To, unsigned long long const Size) APT_OVERRIDE + { + return gzread(gz, To, Size); + } + virtual bool InternalReadError() APT_OVERRIDE + { + int err; + char const * const errmsg = gzerror(gz, &err); + if (err != Z_ERRNO) + return filefd->FileFdError("gzread: %s (%d: %s)", _("Read error"), err, errmsg); + return FileFdPrivate::InternalReadError(); + } + virtual char * InternalReadLine(char * To, unsigned long long Size) APT_OVERRIDE + { + return gzgets(gz, To, Size); + } + virtual ssize_t InternalWrite(void const * const From, unsigned long long const Size) APT_OVERRIDE + { + return gzwrite(gz,From,Size); + } + virtual bool InternalWriteError() APT_OVERRIDE + { + int err; + char const * const errmsg = gzerror(gz, &err); + if (err != Z_ERRNO) + return filefd->FileFdError("gzwrite: %s (%d: %s)", _("Write error"), err, errmsg); + return FileFdPrivate::InternalWriteError(); + } + virtual bool InternalSeek(unsigned long long const To) APT_OVERRIDE + { + off_t const res = gzseek(gz, To, SEEK_SET); + if (res != (off_t)To) + return filefd->FileFdError("Unable to seek to %llu", To); + seekpos = To; + buffer.reset(); + return true; + } + virtual bool InternalSkip(unsigned long long Over) APT_OVERRIDE + { + if (Over >= buffer.size()) + { + Over -= buffer.size(); + buffer.reset(); + } + else + { + buffer.bufferstart += Over; + return true; + } + if (Over == 0) + return true; + off_t const res = gzseek(gz, Over, SEEK_CUR); + if (res < 0) + return filefd->FileFdError("Unable to seek ahead %llu",Over); + seekpos = res; + return true; + } + virtual unsigned long long InternalTell() APT_OVERRIDE + { + return gztell(gz) - buffer.size(); + } + virtual unsigned long long InternalSize() APT_OVERRIDE + { + unsigned long long filesize = FileFdPrivate::InternalSize(); + // only check gzsize if we are actually a gzip file, just checking for + // "gz" is not sufficient as uncompressed files could be opened with + // gzopen in "direct" mode as well + if (filesize == 0 || gzdirect(gz)) + return filesize; + + off_t const oldPos = lseek(filefd->iFd, 0, SEEK_CUR); + /* unfortunately zlib.h doesn't provide a gzsize(), so we have to do + * this ourselves; the original (uncompressed) file size is the last 32 + * bits of the file */ + // FIXME: Size for gz-files is limited by 32bit… no largefile support + if (lseek(filefd->iFd, -4, SEEK_END) < 0) + { + filefd->FileFdErrno("lseek","Unable to seek to end of gzipped file"); + return 0; + } + uint32_t size = 0; + if (read(filefd->iFd, &size, 4) != 4) + { + filefd->FileFdErrno("read","Unable to read original size of gzipped file"); + return 0; + } + size = le32toh(size); + + if (lseek(filefd->iFd, oldPos, SEEK_SET) < 0) + { + filefd->FileFdErrno("lseek","Unable to seek in gzipped file"); + return 0; + } + return size; + } + virtual bool InternalClose(std::string const &FileName) APT_OVERRIDE + { + if (gz == nullptr) + return true; + int const e = gzclose(gz); + gz = nullptr; + // gzdclose() on empty files always fails with "buffer error" here, ignore that + if (e != 0 && e != Z_BUF_ERROR) + return _error->Errno("close",_("Problem closing the gzip file %s"), FileName.c_str()); + return true; + } + + explicit GzipFileFdPrivate(FileFd * const filefd) : FileFdPrivate(filefd), gz(nullptr) {} + virtual ~GzipFileFdPrivate() { InternalClose(""); } +#endif +}; + /*}}}*/ +class APT_HIDDEN Bz2FileFdPrivate: public FileFdPrivate { /*{{{*/ +#ifdef HAVE_BZ2 + BZFILE* bz2; +public: + virtual bool InternalOpen(int const iFd, unsigned int const Mode) APT_OVERRIDE + { + if ((Mode & FileFd::ReadWrite) == FileFd::ReadWrite) + bz2 = BZ2_bzdopen(iFd, "r+"); + else if ((Mode & FileFd::WriteOnly) == FileFd::WriteOnly) + bz2 = BZ2_bzdopen(iFd, "w"); + else + bz2 = BZ2_bzdopen(iFd, "r"); + filefd->Flags |= FileFd::Compressed; + return bz2 != nullptr; + } + virtual ssize_t InternalUnbufferedRead(void * const To, unsigned long long const Size) APT_OVERRIDE + { + return BZ2_bzread(bz2, To, Size); + } + virtual bool InternalReadError() APT_OVERRIDE + { + int err; + char const * const errmsg = BZ2_bzerror(bz2, &err); + if (err != BZ_IO_ERROR) + return filefd->FileFdError("BZ2_bzread: %s %s (%d: %s)", filefd->FileName.c_str(), _("Read error"), err, errmsg); + return FileFdPrivate::InternalReadError(); + } + virtual ssize_t InternalWrite(void const * const From, unsigned long long const Size) APT_OVERRIDE + { + return BZ2_bzwrite(bz2, (void*)From, Size); + } + virtual bool InternalWriteError() APT_OVERRIDE + { + int err; + char const * const errmsg = BZ2_bzerror(bz2, &err); + if (err != BZ_IO_ERROR) + return filefd->FileFdError("BZ2_bzwrite: %s %s (%d: %s)", filefd->FileName.c_str(), _("Write error"), err, errmsg); + return FileFdPrivate::InternalWriteError(); + } + virtual bool InternalStream() const APT_OVERRIDE { return true; } + virtual bool InternalClose(std::string const &) APT_OVERRIDE + { + if (bz2 == nullptr) + return true; + BZ2_bzclose(bz2); + bz2 = nullptr; + return true; + } + + explicit Bz2FileFdPrivate(FileFd * const filefd) : FileFdPrivate(filefd), bz2(nullptr) {} + virtual ~Bz2FileFdPrivate() { InternalClose(""); } +#endif +}; + /*}}}*/ +class APT_HIDDEN Lz4FileFdPrivate: public FileFdPrivate { /*{{{*/ + static constexpr unsigned long long LZ4_HEADER_SIZE = 19; + static constexpr unsigned long long LZ4_FOOTER_SIZE = 4; +#ifdef HAVE_LZ4 + LZ4F_decompressionContext_t dctx; + LZ4F_compressionContext_t cctx; + LZ4F_errorCode_t res; + FileFd backend; + simple_buffer lz4_buffer; + // Count of bytes that the decompressor expects to read next, or buffer size. + size_t next_to_load = APT_BUFFER_SIZE; +public: + virtual bool InternalOpen(int const iFd, unsigned int const Mode) APT_OVERRIDE + { + if ((Mode & FileFd::ReadWrite) == FileFd::ReadWrite) + return _error->Error("lz4 only supports write or read mode"); + + if ((Mode & FileFd::WriteOnly) == FileFd::WriteOnly) { + res = LZ4F_createCompressionContext(&cctx, LZ4F_VERSION); + lz4_buffer.reset(LZ4F_compressBound(APT_BUFFER_SIZE, nullptr) + + LZ4_HEADER_SIZE + LZ4_FOOTER_SIZE); + } else { + res = LZ4F_createDecompressionContext(&dctx, LZ4F_VERSION); + lz4_buffer.reset(APT_BUFFER_SIZE); + } + + filefd->Flags |= FileFd::Compressed; + + if (LZ4F_isError(res)) + return false; + + unsigned int flags = (Mode & (FileFd::WriteOnly|FileFd::ReadOnly)); + if (backend.OpenDescriptor(iFd, flags, FileFd::None, true) == false) + return false; + + // Write the file header + if ((Mode & FileFd::WriteOnly) == FileFd::WriteOnly) + { + res = LZ4F_compressBegin(cctx, lz4_buffer.buffer, lz4_buffer.buffersize_max, nullptr); + if (LZ4F_isError(res) || backend.Write(lz4_buffer.buffer, res) == false) + return false; + } + + return true; + } + virtual ssize_t InternalUnbufferedRead(void * const To, unsigned long long const Size) APT_OVERRIDE + { + /* Keep reading as long as the compressor still wants to read */ + while (next_to_load) { + // Fill compressed buffer; + if (lz4_buffer.empty()) { + unsigned long long read; + /* Reset - if LZ4 decompressor wants to read more, allocate more */ + lz4_buffer.reset(next_to_load); + if (backend.Read(lz4_buffer.getend(), lz4_buffer.free(), &read) == false) + return -1; + lz4_buffer.bufferend += read; + + /* Expected EOF */ + if (read == 0) { + res = -1; + return filefd->FileFdError("LZ4F: %s %s", + filefd->FileName.c_str(), + _("Unexpected end of file")), -1; + } + } + // Drain compressed buffer as far as possible. + size_t in = lz4_buffer.size(); + size_t out = Size; + + res = LZ4F_decompress(dctx, To, &out, lz4_buffer.get(), &in, nullptr); + if (LZ4F_isError(res)) + return -1; + + next_to_load = res; + lz4_buffer.bufferstart += in; + + if (out != 0) + return out; + } + + return 0; + } + virtual bool InternalReadError() APT_OVERRIDE + { + char const * const errmsg = LZ4F_getErrorName(res); + + return filefd->FileFdError("LZ4F: %s %s (%zu: %s)", filefd->FileName.c_str(), _("Read error"), res, errmsg); + } + virtual ssize_t InternalWrite(void const * const From, unsigned long long const Size) APT_OVERRIDE + { + unsigned long long const towrite = std::min(APT_BUFFER_SIZE, Size); + + res = LZ4F_compressUpdate(cctx, + lz4_buffer.buffer, lz4_buffer.buffersize_max, + From, towrite, nullptr); + + if (LZ4F_isError(res) || backend.Write(lz4_buffer.buffer, res) == false) + return -1; + + return towrite; + } + virtual bool InternalWriteError() APT_OVERRIDE + { + char const * const errmsg = LZ4F_getErrorName(res); + + return filefd->FileFdError("LZ4F: %s %s (%zu: %s)", filefd->FileName.c_str(), _("Write error"), res, errmsg); + } + virtual bool InternalStream() const APT_OVERRIDE { return true; } + + virtual bool InternalFlush() APT_OVERRIDE + { + return backend.Flush(); + } + + virtual bool InternalClose(std::string const &) APT_OVERRIDE + { + /* Reset variables */ + res = 0; + next_to_load = APT_BUFFER_SIZE; + + if (cctx != nullptr) + { + if (filefd->Failed() == false) + { + res = LZ4F_compressEnd(cctx, lz4_buffer.buffer, lz4_buffer.buffersize_max, nullptr); + if (LZ4F_isError(res) || backend.Write(lz4_buffer.buffer, res) == false) + return false; + if (!backend.Flush()) + return false; + } + if (!backend.Close()) + return false; + + res = LZ4F_freeCompressionContext(cctx); + cctx = nullptr; + } + + if (dctx != nullptr) + { + res = LZ4F_freeDecompressionContext(dctx); + dctx = nullptr; + } + if (backend.IsOpen()) + { + backend.Close(); + filefd->iFd = -1; + } + + return LZ4F_isError(res) == false; + } + + explicit Lz4FileFdPrivate(FileFd * const filefd) : FileFdPrivate(filefd), dctx(nullptr), cctx(nullptr) {} + virtual ~Lz4FileFdPrivate() { + InternalClose(""); + } +#endif +}; + /*}}}*/ +class APT_HIDDEN ZstdFileFdPrivate : public FileFdPrivate /*{{{*/ +{ +#ifdef HAVE_ZSTD + ZSTD_DStream *dctx; + ZSTD_CStream *cctx; + size_t res = 0; + FileFd backend; + simple_buffer zstd_buffer; + // Count of bytes that the decompressor expects to read next, or buffer size. + size_t next_to_load = APT_BUFFER_SIZE; + + public: + virtual bool InternalOpen(int const iFd, unsigned int const Mode) APT_OVERRIDE + { + if ((Mode & FileFd::ReadWrite) == FileFd::ReadWrite) + return _error->Error("zstd only supports write or read mode"); + + if ((Mode & FileFd::WriteOnly) == FileFd::WriteOnly) + { + cctx = ZSTD_createCStream(); + res = ZSTD_initCStream(cctx, findLevel(compressor.CompressArgs)); + zstd_buffer.reset(APT_BUFFER_SIZE); + } + else + { + dctx = ZSTD_createDStream(); + res = ZSTD_initDStream(dctx); + zstd_buffer.reset(APT_BUFFER_SIZE); + } + + filefd->Flags |= FileFd::Compressed; + + if (ZSTD_isError(res)) + return false; + + unsigned int flags = (Mode & (FileFd::WriteOnly | FileFd::ReadOnly)); + if (backend.OpenDescriptor(iFd, flags, FileFd::None, true) == false) + return false; + + return true; + } + virtual ssize_t InternalUnbufferedRead(void *const To, unsigned long long const Size) APT_OVERRIDE + { + /* Keep reading as long as the compressor still wants to read */ + while (true) + { + // Fill compressed buffer; + if (zstd_buffer.empty()) + { + unsigned long long read; + /* Reset - if LZ4 decompressor wants to read more, allocate more */ + zstd_buffer.reset(next_to_load); + if (backend.Read(zstd_buffer.getend(), zstd_buffer.free(), &read) == false) + return -1; + zstd_buffer.bufferend += read; + + if (read == 0) + { + /* Expected EOF */ + if (next_to_load == 0) + return 0; + + res = -1; + return filefd->FileFdError("ZSTD: %s %s", + filefd->FileName.c_str(), + _("Unexpected end of file")), + -1; + } + } + // Drain compressed buffer as far as possible. + ZSTD_inBuffer in = { + .src = zstd_buffer.get(), + .size = zstd_buffer.size(), + .pos = 0, + }; + ZSTD_outBuffer out = { + .dst = To, + .size = Size, + .pos = 0, + }; + + next_to_load = res = ZSTD_decompressStream(dctx, &out, &in); + + if (res == 0) + { + res = ZSTD_initDStream(dctx); + } + + if (ZSTD_isError(res)) + return -1; + + zstd_buffer.bufferstart += in.pos; + + if (out.pos != 0) + return out.pos; + } + + return 0; + } + virtual bool InternalReadError() APT_OVERRIDE + { + char const *const errmsg = ZSTD_getErrorName(res); + + return filefd->FileFdError("ZSTD: %s %s (%zu: %s)", filefd->FileName.c_str(), _("Read error"), res, errmsg); + } + virtual ssize_t InternalWrite(void const *const From, unsigned long long const Size) APT_OVERRIDE + { + // Drain compressed buffer as far as possible. + ZSTD_outBuffer out = { + .dst = zstd_buffer.buffer, + .size = zstd_buffer.buffersize_max, + .pos = 0, + }; + ZSTD_inBuffer in = { + .src = From, + .size = Size, + .pos = 0, + }; + + res = ZSTD_compressStream(cctx, &out, &in); + + if (ZSTD_isError(res) || backend.Write(zstd_buffer.buffer, out.pos) == false) + return -1; + + return in.pos; + } + + virtual bool InternalWriteError() APT_OVERRIDE + { + char const *const errmsg = ZSTD_getErrorName(res); + + return filefd->FileFdError("ZSTD: %s %s (%zu: %s)", filefd->FileName.c_str(), _("Write error"), res, errmsg); + } + virtual bool InternalStream() const APT_OVERRIDE { return true; } + + virtual bool InternalFlush() APT_OVERRIDE + { + return backend.Flush(); + } + + virtual bool InternalClose(std::string const &) APT_OVERRIDE + { + /* Reset variables */ + res = 0; + next_to_load = APT_BUFFER_SIZE; + + if (cctx != nullptr) + { + if (filefd->Failed() == false) + { + do + { + ZSTD_outBuffer out = { + .dst = zstd_buffer.buffer, + .size = zstd_buffer.buffersize_max, + .pos = 0, + }; + res = ZSTD_endStream(cctx, &out); + if (ZSTD_isError(res) || backend.Write(zstd_buffer.buffer, out.pos) == false) + return false; + } while (res > 0); + + if (!backend.Flush()) + return false; + } + if (!backend.Close()) + return false; + + res = ZSTD_freeCStream(cctx); + cctx = nullptr; + } + + if (dctx != nullptr) + { + res = ZSTD_freeDStream(dctx); + dctx = nullptr; + } + if (backend.IsOpen()) + { + backend.Close(); + filefd->iFd = -1; + } + + return ZSTD_isError(res) == false; + } + + static uint32_t findLevel(std::vector<std::string> const &Args) + { + for (auto a = Args.rbegin(); a != Args.rend(); ++a) + { + if (a->size() >= 2 && (*a)[0] == '-' && (*a)[1] != '-') + { + auto const level = a->substr(1); + auto const notANumber = level.find_first_not_of("0123456789"); + if (notANumber != std::string::npos) + continue; + + return (uint32_t)stoi(level); + } + } + return 19; + } + + explicit ZstdFileFdPrivate(FileFd *const filefd) : FileFdPrivate(filefd), dctx(nullptr), cctx(nullptr) {} + virtual ~ZstdFileFdPrivate() + { + InternalClose(""); + } +#endif +}; + /*}}}*/ +class APT_HIDDEN LzmaFileFdPrivate: public FileFdPrivate { /*{{{*/ +#ifdef HAVE_LZMA + struct LZMAFILE { + FILE* file; + FileFd * const filefd; + uint8_t buffer[4096]; + lzma_stream stream; + lzma_ret err; + bool eof; + bool compressing; + + explicit LZMAFILE(FileFd * const fd) : file(nullptr), filefd(fd), eof(false), compressing(false) { buffer[0] = '\0'; } + ~LZMAFILE() + { + if (compressing == true && filefd->Failed() == false) + { + size_t constexpr buffersize = sizeof(buffer)/sizeof(buffer[0]); + while(true) + { + stream.avail_out = buffersize; + stream.next_out = buffer; + err = lzma_code(&stream, LZMA_FINISH); + if (err != LZMA_OK && err != LZMA_STREAM_END) + { + _error->Error("~LZMAFILE: Compress finalisation failed"); + break; + } + size_t const n = buffersize - stream.avail_out; + if (n && fwrite(buffer, 1, n, file) != n) + { + _error->Errno("~LZMAFILE",_("Write error")); + break; + } + if (err == LZMA_STREAM_END) + break; + } + } + lzma_end(&stream); + fclose(file); + } + }; + LZMAFILE* lzma; + static uint32_t findXZlevel(std::vector<std::string> const &Args) + { + for (auto a = Args.rbegin(); a != Args.rend(); ++a) + if (a->empty() == false && (*a)[0] == '-' && (*a)[1] != '-') + { + auto const number = a->find_last_of("0123456789"); + if (number == std::string::npos) + continue; + auto const extreme = a->find("e", number); + uint32_t level = (extreme != std::string::npos) ? LZMA_PRESET_EXTREME : 0; + switch ((*a)[number]) + { + case '0': return level | 0; + case '1': return level | 1; + case '2': return level | 2; + case '3': return level | 3; + case '4': return level | 4; + case '5': return level | 5; + case '6': return level | 6; + case '7': return level | 7; + case '8': return level | 8; + case '9': return level | 9; + } + } + return 6; + } +public: + virtual bool InternalOpen(int const iFd, unsigned int const Mode) APT_OVERRIDE + { + if ((Mode & FileFd::ReadWrite) == FileFd::ReadWrite) + return filefd->FileFdError("ReadWrite mode is not supported for lzma/xz files %s", filefd->FileName.c_str()); + + if (lzma == nullptr) + lzma = new LzmaFileFdPrivate::LZMAFILE(filefd); + if ((Mode & FileFd::WriteOnly) == FileFd::WriteOnly) + lzma->file = fdopen(iFd, "w"); + else + lzma->file = fdopen(iFd, "r"); + filefd->Flags |= FileFd::Compressed; + if (lzma->file == nullptr) + return false; + + lzma_stream tmp_stream = LZMA_STREAM_INIT; + lzma->stream = tmp_stream; + + if ((Mode & FileFd::WriteOnly) == FileFd::WriteOnly) + { + uint32_t const xzlevel = findXZlevel(compressor.CompressArgs); + if (compressor.Name == "xz") + { + if (lzma_easy_encoder(&lzma->stream, xzlevel, LZMA_CHECK_CRC64) != LZMA_OK) + return false; + } + else + { + lzma_options_lzma options; + lzma_lzma_preset(&options, xzlevel); + if (lzma_alone_encoder(&lzma->stream, &options) != LZMA_OK) + return false; + } + lzma->compressing = true; + } + else + { + uint64_t constexpr memlimit = 1024 * 1024 * 500; + if (lzma_auto_decoder(&lzma->stream, memlimit, 0) != LZMA_OK) + return false; + lzma->compressing = false; + } + return true; + } + virtual ssize_t InternalUnbufferedRead(void * const To, unsigned long long const Size) APT_OVERRIDE + { + ssize_t Res; + if (lzma->eof == true) + return 0; + + lzma->stream.next_out = (uint8_t *) To; + lzma->stream.avail_out = Size; + if (lzma->stream.avail_in == 0) + { + lzma->stream.next_in = lzma->buffer; + lzma->stream.avail_in = fread(lzma->buffer, 1, sizeof(lzma->buffer)/sizeof(lzma->buffer[0]), lzma->file); + } + lzma->err = lzma_code(&lzma->stream, LZMA_RUN); + if (lzma->err == LZMA_STREAM_END) + { + lzma->eof = true; + Res = Size - lzma->stream.avail_out; + } + else if (lzma->err != LZMA_OK) + { + Res = -1; + errno = 0; + } + else + { + Res = Size - lzma->stream.avail_out; + if (Res == 0) + { + // lzma run was okay, but produced no output… + Res = -1; + errno = EINTR; + } + } + return Res; + } + virtual bool InternalReadError() APT_OVERRIDE + { + return filefd->FileFdError("lzma_read: %s (%d)", _("Read error"), lzma->err); + } + virtual ssize_t InternalWrite(void const * const From, unsigned long long const Size) APT_OVERRIDE + { + ssize_t Res; + lzma->stream.next_in = (uint8_t *)From; + lzma->stream.avail_in = Size; + lzma->stream.next_out = lzma->buffer; + lzma->stream.avail_out = sizeof(lzma->buffer)/sizeof(lzma->buffer[0]); + lzma->err = lzma_code(&lzma->stream, LZMA_RUN); + if (lzma->err != LZMA_OK) + return -1; + size_t const n = sizeof(lzma->buffer)/sizeof(lzma->buffer[0]) - lzma->stream.avail_out; + size_t const m = (n == 0) ? 0 : fwrite(lzma->buffer, 1, n, lzma->file); + if (m != n) + { + Res = -1; + errno = 0; + } + else + { + Res = Size - lzma->stream.avail_in; + if (Res == 0) + { + // lzma run was okay, but produced no output… + Res = -1; + errno = EINTR; + } + } + return Res; + } + virtual bool InternalWriteError() APT_OVERRIDE + { + return filefd->FileFdError("lzma_write: %s (%d)", _("Write error"), lzma->err); + } + virtual bool InternalStream() const APT_OVERRIDE { return true; } + virtual bool InternalClose(std::string const &) APT_OVERRIDE + { + delete lzma; + lzma = nullptr; + return true; + } + + explicit LzmaFileFdPrivate(FileFd * const filefd) : FileFdPrivate(filefd), lzma(nullptr) {} + virtual ~LzmaFileFdPrivate() { InternalClose(""); } +#endif +}; + /*}}}*/ +class APT_HIDDEN PipedFileFdPrivate: public FileFdPrivate /*{{{*/ +/* if we don't have a specific class dealing with library calls, we (un)compress + by executing a specified binary and pipe in/out what we need */ +{ +public: + virtual bool InternalOpen(int const, unsigned int const Mode) APT_OVERRIDE + { + // collect zombies here in case we reopen + if (compressor_pid > 0) + ExecWait(compressor_pid, "FileFdCompressor", true); + + if ((Mode & FileFd::ReadWrite) == FileFd::ReadWrite) + return filefd->FileFdError("ReadWrite mode is not supported for file %s", filefd->FileName.c_str()); + if (compressor.Binary == "false") + return filefd->FileFdError("libapt has inbuilt support for the %s compression," + " but was forced to ignore it in favor of an external binary – which isn't installed.", compressor.Name.c_str()); + + bool const Comp = (Mode & FileFd::WriteOnly) == FileFd::WriteOnly; + if (Comp == false && filefd->iFd != -1) + { + // Handle 'decompression' of empty files + struct stat Buf; + if (fstat(filefd->iFd, &Buf) != 0) + return filefd->FileFdErrno("fstat", "Could not stat fd %d for file %s", filefd->iFd, filefd->FileName.c_str()); + if (Buf.st_size == 0 && S_ISFIFO(Buf.st_mode) == false) + return true; + + // We don't need the file open - instead let the compressor open it + // as he properly knows better how to efficiently read from 'his' file + if (filefd->FileName.empty() == false) + { + close(filefd->iFd); + filefd->iFd = -1; + } + } + + // Create a data pipe + int Pipe[2] = {-1,-1}; + if (pipe(Pipe) != 0) + return filefd->FileFdErrno("pipe",_("Failed to create subprocess IPC")); + for (int J = 0; J != 2; J++) + SetCloseExec(Pipe[J],true); + + compressed_fd = filefd->iFd; + set_is_pipe(true); + + if (Comp == true) + filefd->iFd = Pipe[1]; + else + filefd->iFd = Pipe[0]; + + // The child.. + compressor_pid = ExecFork(); + if (compressor_pid == 0) + { + if (Comp == true) + { + dup2(compressed_fd,STDOUT_FILENO); + dup2(Pipe[0],STDIN_FILENO); + } + else + { + if (compressed_fd != -1) + dup2(compressed_fd,STDIN_FILENO); + dup2(Pipe[1],STDOUT_FILENO); + } + int const nullfd = open("/dev/null", O_WRONLY); + if (nullfd != -1) + { + dup2(nullfd,STDERR_FILENO); + close(nullfd); + } + + SetCloseExec(STDOUT_FILENO,false); + SetCloseExec(STDIN_FILENO,false); + + std::vector<char const*> Args; + Args.push_back(compressor.Binary.c_str()); + std::vector<std::string> const * const addArgs = + (Comp == true) ? &(compressor.CompressArgs) : &(compressor.UncompressArgs); + for (std::vector<std::string>::const_iterator a = addArgs->begin(); + a != addArgs->end(); ++a) + Args.push_back(a->c_str()); + if (Comp == false && filefd->FileName.empty() == false) + { + // commands not needing arguments, do not need to be told about using standard output + // in reality, only testcases with tools like cat, rev, rot13, … are able to trigger this + if (compressor.CompressArgs.empty() == false && compressor.UncompressArgs.empty() == false) + Args.push_back("--stdout"); + if (filefd->TemporaryFileName.empty() == false) + Args.push_back(filefd->TemporaryFileName.c_str()); + else + Args.push_back(filefd->FileName.c_str()); + } + Args.push_back(NULL); + + execvp(Args[0],(char **)&Args[0]); + cerr << _("Failed to exec compressor ") << Args[0] << endl; + _exit(100); + } + if (Comp == true) + close(Pipe[0]); + else + close(Pipe[1]); + + return true; + } + virtual ssize_t InternalUnbufferedRead(void * const To, unsigned long long const Size) APT_OVERRIDE + { + return read(filefd->iFd, To, Size); + } + virtual ssize_t InternalWrite(void const * const From, unsigned long long const Size) APT_OVERRIDE + { + return write(filefd->iFd, From, Size); + } + virtual bool InternalClose(std::string const &) APT_OVERRIDE + { + bool Ret = true; + if (filefd->iFd != -1) + { + close(filefd->iFd); + filefd->iFd = -1; + } + if (compressor_pid > 0) + Ret &= ExecWait(compressor_pid, "FileFdCompressor", true); + compressor_pid = -1; + return Ret; + } + explicit PipedFileFdPrivate(FileFd * const filefd) : FileFdPrivate(filefd) {} + virtual ~PipedFileFdPrivate() { InternalClose(""); } +}; + /*}}}*/ +class APT_HIDDEN DirectFileFdPrivate: public FileFdPrivate /*{{{*/ +{ +public: + virtual bool InternalOpen(int const, unsigned int const) APT_OVERRIDE { return true; } + virtual ssize_t InternalUnbufferedRead(void * const To, unsigned long long const Size) APT_OVERRIDE + { + return read(filefd->iFd, To, Size); + } + virtual ssize_t InternalWrite(void const * const From, unsigned long long const Size) APT_OVERRIDE + { + // files opened read+write are strange and only really "supported" for direct files + if (buffer.size() != 0) + { + lseek(filefd->iFd, -buffer.size(), SEEK_CUR); + buffer.reset(); + } + return write(filefd->iFd, From, Size); + } + virtual bool InternalSeek(unsigned long long const To) APT_OVERRIDE + { + off_t const res = lseek(filefd->iFd, To, SEEK_SET); + if (res != (off_t)To) + return filefd->FileFdError("Unable to seek to %llu", To); + seekpos = To; + buffer.reset(); + return true; + } + virtual bool InternalSkip(unsigned long long Over) APT_OVERRIDE + { + if (Over >= buffer.size()) + { + Over -= buffer.size(); + buffer.reset(); + } + else + { + buffer.bufferstart += Over; + return true; + } + if (Over == 0) + return true; + off_t const res = lseek(filefd->iFd, Over, SEEK_CUR); + if (res < 0) + return filefd->FileFdError("Unable to seek ahead %llu",Over); + seekpos = res; + return true; + } + virtual bool InternalTruncate(unsigned long long const To) APT_OVERRIDE + { + if (buffer.size() != 0) + { + unsigned long long const seekpos = lseek(filefd->iFd, 0, SEEK_CUR); + if ((seekpos - buffer.size()) >= To) + buffer.reset(); + else if (seekpos >= To) + buffer.bufferend = (To - seekpos) + buffer.bufferstart; + else + buffer.reset(); + } + if (ftruncate(filefd->iFd, To) != 0) + return filefd->FileFdError("Unable to truncate to %llu",To); + return true; + } + virtual unsigned long long InternalTell() APT_OVERRIDE + { + return lseek(filefd->iFd,0,SEEK_CUR) - buffer.size(); + } + virtual unsigned long long InternalSize() APT_OVERRIDE + { + return filefd->FileSize(); + } + virtual bool InternalClose(std::string const &) APT_OVERRIDE { return true; } + virtual bool InternalAlwaysAutoClose() const APT_OVERRIDE { return false; } + + explicit DirectFileFdPrivate(FileFd * const filefd) : FileFdPrivate(filefd) {} + virtual ~DirectFileFdPrivate() { InternalClose(""); } +}; + /*}}}*/ +// FileFd Constructors /*{{{*/ +FileFd::FileFd(std::string FileName,unsigned int const Mode,unsigned long AccessMode) : iFd(-1), Flags(0), d(NULL) +{ + Open(FileName,Mode, None, AccessMode); +} +FileFd::FileFd(std::string FileName,unsigned int const Mode, CompressMode Compress, unsigned long AccessMode) : iFd(-1), Flags(0), d(NULL) +{ + Open(FileName,Mode, Compress, AccessMode); +} +FileFd::FileFd() : iFd(-1), Flags(AutoClose), d(NULL) {} +FileFd::FileFd(int const Fd, unsigned int const Mode, CompressMode Compress) : iFd(-1), Flags(0), d(NULL) +{ + OpenDescriptor(Fd, Mode, Compress); +} +FileFd::FileFd(int const Fd, bool const AutoClose) : iFd(-1), Flags(0), d(NULL) +{ + OpenDescriptor(Fd, ReadWrite, None, AutoClose); +} + /*}}}*/ +// FileFd::Open - Open a file /*{{{*/ +// --------------------------------------------------------------------- +/* The most commonly used open mode combinations are given with Mode */ +bool FileFd::Open(string FileName,unsigned int const Mode,CompressMode Compress, unsigned long const AccessMode) +{ + if (Mode == ReadOnlyGzip) + return Open(FileName, ReadOnly, Gzip, AccessMode); + + if (Compress == Auto && (Mode & WriteOnly) == WriteOnly) + return FileFdError("Autodetection on %s only works in ReadOnly openmode!", FileName.c_str()); + + std::vector<APT::Configuration::Compressor> const compressors = APT::Configuration::getCompressors(); + std::vector<APT::Configuration::Compressor>::const_iterator compressor = compressors.begin(); + if (Compress == Auto) + { + for (; compressor != compressors.end(); ++compressor) + { + std::string file = FileName + compressor->Extension; + if (FileExists(file) == false) + continue; + FileName = file; + break; + } + } + else if (Compress == Extension) + { + std::string::size_type const found = FileName.find_last_of('.'); + std::string ext; + if (found != std::string::npos) + { + ext = FileName.substr(found); + if (ext == ".new" || ext == ".bak") + { + std::string::size_type const found2 = FileName.find_last_of('.', found - 1); + if (found2 != std::string::npos) + ext = FileName.substr(found2, found - found2); + else + ext.clear(); + } + } + for (; compressor != compressors.end(); ++compressor) + if (ext == compressor->Extension) + break; + // no matching extension - assume uncompressed (imagine files like 'example.org_Packages') + if (compressor == compressors.end()) + for (compressor = compressors.begin(); compressor != compressors.end(); ++compressor) + if (compressor->Name == ".") + break; + } + else + { + std::string name; + switch (Compress) + { + case None: name = "."; break; + case Gzip: name = "gzip"; break; + case Bzip2: name = "bzip2"; break; + case Lzma: name = "lzma"; break; + case Xz: name = "xz"; break; + case Lz4: name = "lz4"; break; + case Zstd: name = "zstd"; break; + case Auto: + case Extension: + // Unreachable + return FileFdError("Opening File %s in None, Auto or Extension should be already handled?!?", FileName.c_str()); + } + for (; compressor != compressors.end(); ++compressor) + if (compressor->Name == name) + break; + if (compressor == compressors.end()) + return FileFdError("Can't find a configured compressor %s for file %s", name.c_str(), FileName.c_str()); + } + + if (compressor == compressors.end()) + return FileFdError("Can't find a match for specified compressor mode for file %s", FileName.c_str()); + return Open(FileName, Mode, *compressor, AccessMode); +} +bool FileFd::Open(string FileName,unsigned int const Mode,APT::Configuration::Compressor const &compressor, unsigned long const AccessMode) +{ + Close(); + Flags = AutoClose; + + if ((Mode & WriteOnly) != WriteOnly && (Mode & (Atomic | Create | Empty | Exclusive)) != 0) + return FileFdError("ReadOnly mode for %s doesn't accept additional flags!", FileName.c_str()); + if ((Mode & ReadWrite) == 0) + return FileFdError("No openmode provided in FileFd::Open for %s", FileName.c_str()); + + unsigned int OpenMode = Mode; + if (FileName == "/dev/null") + OpenMode = OpenMode & ~(Atomic | Exclusive | Create | Empty); + + if ((OpenMode & Atomic) == Atomic) + { + Flags |= Replace; + } + else if ((OpenMode & (Exclusive | Create)) == (Exclusive | Create)) + { + // for atomic, this will be done by rename in Close() + RemoveFile("FileFd::Open", FileName); + } + if ((OpenMode & Empty) == Empty) + { + struct stat Buf; + if (lstat(FileName.c_str(),&Buf) == 0 && S_ISLNK(Buf.st_mode)) + RemoveFile("FileFd::Open", FileName); + } + + int fileflags = 0; + #define if_FLAGGED_SET(FLAG, MODE) if ((OpenMode & FLAG) == FLAG) fileflags |= MODE + if_FLAGGED_SET(ReadWrite, O_RDWR); + else if_FLAGGED_SET(ReadOnly, O_RDONLY); + else if_FLAGGED_SET(WriteOnly, O_WRONLY); + + if_FLAGGED_SET(Create, O_CREAT); + if_FLAGGED_SET(Empty, O_TRUNC); + if_FLAGGED_SET(Exclusive, O_EXCL); + #undef if_FLAGGED_SET + + if ((OpenMode & Atomic) == Atomic) + { + char *name = strdup((FileName + ".XXXXXX").c_str()); + + if((iFd = mkstemp(name)) == -1) + { + free(name); + return FileFdErrno("mkstemp", "Could not create temporary file for %s", FileName.c_str()); + } + + TemporaryFileName = string(name); + free(name); + + // umask() will always set the umask and return the previous value, so + // we first set the umask and then reset it to the old value + mode_t const CurrentUmask = umask(0); + umask(CurrentUmask); + // calculate the actual file permissions (just like open/creat) + mode_t const FilePermissions = (AccessMode & ~CurrentUmask); + + if(fchmod(iFd, FilePermissions) == -1) + return FileFdErrno("fchmod", "Could not change permissions for temporary file %s", TemporaryFileName.c_str()); + } + else + iFd = open(FileName.c_str(), fileflags, AccessMode); + + this->FileName = FileName; + if (iFd == -1 || OpenInternDescriptor(OpenMode, compressor) == false) + { + if (iFd != -1) + { + close (iFd); + iFd = -1; + } + return FileFdErrno("open",_("Could not open file %s"), FileName.c_str()); + } + + SetCloseExec(iFd,true); + return true; +} + /*}}}*/ +// FileFd::OpenDescriptor - Open a filedescriptor /*{{{*/ +bool FileFd::OpenDescriptor(int Fd, unsigned int const Mode, CompressMode Compress, bool AutoClose) +{ + std::vector<APT::Configuration::Compressor> const compressors = APT::Configuration::getCompressors(); + std::vector<APT::Configuration::Compressor>::const_iterator compressor = compressors.begin(); + std::string name; + + // compat with the old API + if (Mode == ReadOnlyGzip && Compress == None) + Compress = Gzip; + + switch (Compress) + { + case None: name = "."; break; + case Gzip: name = "gzip"; break; + case Bzip2: name = "bzip2"; break; + case Lzma: name = "lzma"; break; + case Xz: name = "xz"; break; + case Lz4: name = "lz4"; break; + case Zstd: name = "zstd"; break; + case Auto: + case Extension: + if (AutoClose == true && Fd != -1) + close(Fd); + return FileFdError("Opening Fd %d in Auto or Extension compression mode is not supported", Fd); + } + for (; compressor != compressors.end(); ++compressor) + if (compressor->Name == name) + break; + if (compressor == compressors.end()) + { + if (AutoClose == true && Fd != -1) + close(Fd); + return FileFdError("Can't find a configured compressor %s for file %s", name.c_str(), FileName.c_str()); + } + return OpenDescriptor(Fd, Mode, *compressor, AutoClose); +} +bool FileFd::OpenDescriptor(int Fd, unsigned int const Mode, APT::Configuration::Compressor const &compressor, bool AutoClose) +{ + Close(); + Flags = (AutoClose) ? FileFd::AutoClose : 0; + iFd = Fd; + this->FileName = ""; + if (OpenInternDescriptor(Mode, compressor) == false) + { + if (iFd != -1 && ( + (Flags & Compressed) == Compressed || + AutoClose == true)) + { + close (iFd); + iFd = -1; + } + return FileFdError(_("Could not open file descriptor %d"), Fd); + } + return true; +} +bool FileFd::OpenInternDescriptor(unsigned int const Mode, APT::Configuration::Compressor const &compressor) +{ + if (iFd == -1) + return false; + + if (d != nullptr) + d->InternalClose(FileName); + + if (d == nullptr) + { + if (false) + /* dummy so that the rest can be 'else if's */; +#define APT_COMPRESS_INIT(NAME, CONSTRUCTOR) \ + else if (compressor.Name == NAME) \ + d = new CONSTRUCTOR(this) +#ifdef HAVE_ZLIB + APT_COMPRESS_INIT("gzip", GzipFileFdPrivate); +#endif +#ifdef HAVE_BZ2 + APT_COMPRESS_INIT("bzip2", Bz2FileFdPrivate); +#endif +#ifdef HAVE_LZMA + APT_COMPRESS_INIT("xz", LzmaFileFdPrivate); + APT_COMPRESS_INIT("lzma", LzmaFileFdPrivate); +#endif +#ifdef HAVE_LZ4 + APT_COMPRESS_INIT("lz4", Lz4FileFdPrivate); +#endif +#ifdef HAVE_ZSTD + APT_COMPRESS_INIT("zstd", ZstdFileFdPrivate); +#endif +#undef APT_COMPRESS_INIT + else if (compressor.Name == "." || compressor.Binary.empty() == true) + d = new DirectFileFdPrivate(this); + else + d = new PipedFileFdPrivate(this); + + if (Mode & BufferedWrite) + d = new BufferedWriteFileFdPrivate(d); + + d->set_openmode(Mode); + d->set_compressor(compressor); + if ((Flags & AutoClose) != AutoClose && d->InternalAlwaysAutoClose()) + { + // Need to duplicate fd here or gz/bz2 close for cleanup will close the fd as well + int const internFd = dup(iFd); + if (internFd == -1) + return FileFdErrno("OpenInternDescriptor", _("Could not open file descriptor %d"), iFd); + iFd = internFd; + } + } + return d->InternalOpen(iFd, Mode); +} + /*}}}*/ +// FileFd::~File - Closes the file /*{{{*/ +// --------------------------------------------------------------------- +/* If the proper modes are selected then we close the Fd and possibly + unlink the file on error. */ +FileFd::~FileFd() +{ + Close(); + if (d != NULL) + d->InternalClose(FileName); + delete d; + d = NULL; +} + /*}}}*/ +// FileFd::Read - Read a bit of the file /*{{{*/ +// --------------------------------------------------------------------- +/* We are careful to handle interruption by a signal while reading + gracefully. */ +bool FileFd::Read(void *To,unsigned long long Size,unsigned long long *Actual) +{ + if (d == nullptr || Failed()) + return false; + ssize_t Res = 1; + errno = 0; + if (Actual != 0) + *Actual = 0; + *((char *)To) = '\0'; + while (Res > 0 && Size > 0) + { + Res = d->InternalRead(To, Size); + + if (Res < 0) + { + if (errno == EINTR) + { + // trick the while-loop into running again + Res = 1; + errno = 0; + continue; + } + return d->InternalReadError(); + } + + To = (char *)To + Res; + Size -= Res; + if (d != NULL) + d->set_seekpos(d->get_seekpos() + Res); + if (Actual != 0) + *Actual += Res; + } + + if (Size == 0) + return true; + + // Eof handling + if (Actual != 0) + { + Flags |= HitEof; + return true; + } + + return FileFdError(_("read, still have %llu to read but none left"), Size); +} +bool FileFd::Read(int const Fd, void *To, unsigned long long Size, unsigned long long * const Actual) +{ + ssize_t Res = 1; + errno = 0; + if (Actual != nullptr) + *Actual = 0; + *static_cast<char *>(To) = '\0'; + while (Res > 0 && Size > 0) + { + Res = read(Fd, To, Size); + if (Res < 0) + { + if (errno == EINTR) + { + Res = 1; + errno = 0; + continue; + } + return _error->Errno("read", _("Read error")); + } + To = static_cast<char *>(To) + Res; + Size -= Res; + if (Actual != 0) + *Actual += Res; + } + if (Size == 0) + return true; + if (Actual != nullptr) + return true; + return _error->Error(_("read, still have %llu to read but none left"), Size); +} + /*}}}*/ +// FileFd::ReadLine - Read a complete line from the file /*{{{*/ +char* FileFd::ReadLine(char *To, unsigned long long const Size) +{ + *To = '\0'; + if (d == nullptr || Failed()) + return nullptr; + return d->InternalReadLine(To, Size); +} +bool FileFd::ReadLine(std::string &To) +{ + To.clear(); + if (d == nullptr || Failed()) + return false; + constexpr size_t buflen = 4096; + char buffer[buflen]; + size_t len; + do + { + if (d->InternalReadLine(buffer, buflen) == nullptr) + return false; + len = strlen(buffer); + To.append(buffer, len); + } while (len == buflen - 1 && buffer[len - 2] != '\n'); + // remove the newline at the end + auto const i = To.find_last_not_of("\r\n"); + if (i == std::string::npos) + To.clear(); + else + To.erase(i + 1); + return true; +} + /*}}}*/ +// FileFd::Flush - Flush the file /*{{{*/ +bool FileFd::Flush() +{ + if (Failed()) + return false; + if (d == nullptr) + return true; + + return d->InternalFlush(); +} + /*}}}*/ +// FileFd::Write - Write to the file /*{{{*/ +bool FileFd::Write(const void *From,unsigned long long Size) +{ + if (d == nullptr || Failed()) + return false; + ssize_t Res = 1; + errno = 0; + while (Res > 0 && Size > 0) + { + Res = d->InternalWrite(From, Size); + + if (Res < 0) + { + if (errno == EINTR) + { + // trick the while-loop into running again + Res = 1; + errno = 0; + continue; + } + return d->InternalWriteError(); + } + + From = (char const *)From + Res; + Size -= Res; + if (d != NULL) + d->set_seekpos(d->get_seekpos() + Res); + } + + if (Size == 0) + return true; + + return FileFdError(_("write, still have %llu to write but couldn't"), Size); +} +bool FileFd::Write(int Fd, const void *From, unsigned long long Size) +{ + ssize_t Res = 1; + errno = 0; + while (Res > 0 && Size > 0) + { + Res = write(Fd,From,Size); + if (Res < 0 && errno == EINTR) + continue; + if (Res < 0) + return _error->Errno("write",_("Write error")); + + From = (char const *)From + Res; + Size -= Res; + } + + if (Size == 0) + return true; + + return _error->Error(_("write, still have %llu to write but couldn't"), Size); +} + /*}}}*/ +// FileFd::Seek - Seek in the file /*{{{*/ +bool FileFd::Seek(unsigned long long To) +{ + if (d == nullptr || Failed()) + return false; + Flags &= ~HitEof; + return d->InternalSeek(To); +} + /*}}}*/ +// FileFd::Skip - Skip over data in the file /*{{{*/ +bool FileFd::Skip(unsigned long long Over) +{ + if (d == nullptr || Failed()) + return false; + return d->InternalSkip(Over); +} + /*}}}*/ +// FileFd::Truncate - Truncate the file /*{{{*/ +bool FileFd::Truncate(unsigned long long To) +{ + if (d == nullptr || Failed()) + return false; + // truncating /dev/null is always successful - as we get an error otherwise + if (To == 0 && FileName == "/dev/null") + return true; + return d->InternalTruncate(To); +} + /*}}}*/ +// FileFd::Tell - Current seek position /*{{{*/ +// --------------------------------------------------------------------- +/* */ +unsigned long long FileFd::Tell() +{ + if (d == nullptr || Failed()) + return false; + off_t const Res = d->InternalTell(); + if (Res == (off_t)-1) + FileFdErrno("lseek","Failed to determine the current file position"); + d->set_seekpos(Res); + return Res; +} + /*}}}*/ +static bool StatFileFd(char const * const msg, int const iFd, std::string const &FileName, struct stat &Buf, FileFdPrivate * const d) /*{{{*/ +{ + bool ispipe = (d != NULL && d->get_is_pipe() == true); + if (ispipe == false) + { + if (fstat(iFd,&Buf) != 0) + // higher-level code will generate more meaningful messages, + // even translated this would be meaningless for users + return _error->Errno("fstat", "Unable to determine %s for fd %i", msg, iFd); + if (FileName.empty() == false) + ispipe = S_ISFIFO(Buf.st_mode); + } + + // for compressor pipes st_size is undefined and at 'best' zero + if (ispipe == true) + { + // we set it here, too, as we get the info here for free + // in theory the Open-methods should take care of it already + if (d != NULL) + d->set_is_pipe(true); + if (stat(FileName.c_str(), &Buf) != 0) + return _error->Errno("fstat", "Unable to determine %s for file %s", msg, FileName.c_str()); + } + return true; +} + /*}}}*/ +// FileFd::FileSize - Return the size of the file /*{{{*/ +unsigned long long FileFd::FileSize() +{ + struct stat Buf; + if (StatFileFd("file size", iFd, FileName, Buf, d) == false) + { + Flags |= Fail; + return 0; + } + return Buf.st_size; +} + /*}}}*/ +// FileFd::ModificationTime - Return the time of last touch /*{{{*/ +time_t FileFd::ModificationTime() +{ + struct stat Buf; + if (StatFileFd("modification time", iFd, FileName, Buf, d) == false) + { + Flags |= Fail; + return 0; + } + return Buf.st_mtime; +} + /*}}}*/ +// FileFd::Size - Return the size of the content in the file /*{{{*/ +unsigned long long FileFd::Size() +{ + if (d == nullptr) + return 0; + return d->InternalSize(); +} + /*}}}*/ +// FileFd::Close - Close the file if the close flag is set /*{{{*/ +// --------------------------------------------------------------------- +/* */ +bool FileFd::Close() +{ + if (Failed() == false && Flush() == false) + return false; + if (iFd == -1) + return true; + + bool Res = true; + if ((Flags & AutoClose) == AutoClose) + { + if ((Flags & Compressed) != Compressed && iFd > 0 && close(iFd) != 0) + Res &= _error->Errno("close",_("Problem closing the file %s"), FileName.c_str()); + } + + if (d != NULL) + { + Res &= d->InternalClose(FileName); + delete d; + d = NULL; + } + + if ((Flags & Replace) == Replace) { + if (Failed() == false && rename(TemporaryFileName.c_str(), FileName.c_str()) != 0) + Res &= _error->Errno("rename",_("Problem renaming the file %s to %s"), TemporaryFileName.c_str(), FileName.c_str()); + + FileName = TemporaryFileName; // for the unlink() below. + TemporaryFileName.clear(); + } + + iFd = -1; + + if ((Flags & Fail) == Fail && (Flags & DelOnFail) == DelOnFail && + FileName.empty() == false) + Res &= RemoveFile("FileFd::Close", FileName); + + if (Res == false) + Flags |= Fail; + return Res; +} + /*}}}*/ +// FileFd::Sync - Sync the file /*{{{*/ +// --------------------------------------------------------------------- +/* */ +bool FileFd::Sync() +{ + if (fsync(iFd) != 0) + return FileFdErrno("sync",_("Problem syncing the file")); + return true; +} + /*}}}*/ +// FileFd::FileFdErrno - set Fail and call _error->Errno *{{{*/ +bool FileFd::FileFdErrno(const char *Function, const char *Description,...) +{ + Flags |= Fail; + va_list args; + size_t msgSize = 400; + int const errsv = errno; + bool retry; + do { + va_start(args,Description); + retry = _error->InsertErrno(GlobalError::ERROR, Function, Description, args, errsv, msgSize); + va_end(args); + } while (retry); + return false; +} + /*}}}*/ +// FileFd::FileFdError - set Fail and call _error->Error *{{{*/ +bool FileFd::FileFdError(const char *Description,...) { + Flags |= Fail; + va_list args; + size_t msgSize = 400; + bool retry; + do { + va_start(args,Description); + retry = _error->Insert(GlobalError::ERROR, Description, args, msgSize); + va_end(args); + } while (retry); + return false; +} + /*}}}*/ +// Glob - wrapper around "glob()" /*{{{*/ +std::vector<std::string> Glob(std::string const &pattern, int flags) +{ + std::vector<std::string> result; + glob_t globbuf; + int glob_res; + unsigned int i; + + glob_res = glob(pattern.c_str(), flags, NULL, &globbuf); + + if (glob_res != 0) + { + if(glob_res != GLOB_NOMATCH) { + _error->Errno("glob", "Problem with glob"); + return result; + } + } + + // append results + for(i=0;i<globbuf.gl_pathc;i++) + result.push_back(string(globbuf.gl_pathv[i])); + + globfree(&globbuf); + return result; +} + /*}}}*/ +static std::string APT_NONNULL(1) GetTempDirEnv(char const * const env) /*{{{*/ +{ + const char *tmpdir = getenv(env); + +#ifdef P_tmpdir + if (!tmpdir) + tmpdir = P_tmpdir; +#endif + + struct stat st; + if (!tmpdir || strlen(tmpdir) == 0 || // tmpdir is set + stat(tmpdir, &st) != 0 || (st.st_mode & S_IFDIR) == 0) // exists and is directory + tmpdir = "/tmp"; + else if (geteuid() != 0 && // root can do everything anyway + faccessat(AT_FDCWD, tmpdir, R_OK | W_OK | X_OK, AT_EACCESS) != 0) // current user has rwx access to directory + tmpdir = "/tmp"; + + return string(tmpdir); +} + /*}}}*/ +std::string GetTempDir() /*{{{*/ +{ + return GetTempDirEnv("TMPDIR"); +} +std::string GetTempDir(std::string const &User) +{ + // no need/possibility to drop privs + if(getuid() != 0 || User.empty() || User == "root") + return GetTempDir(); + + struct passwd const * const pw = getpwnam(User.c_str()); + if (pw == NULL) + return GetTempDir(); + + gid_t const old_euid = geteuid(); + gid_t const old_egid = getegid(); + if (setegid(pw->pw_gid) != 0) + _error->Errno("setegid", "setegid %u failed", pw->pw_gid); + if (seteuid(pw->pw_uid) != 0) + _error->Errno("seteuid", "seteuid %u failed", pw->pw_uid); + + std::string const tmp = GetTempDir(); + + if (seteuid(old_euid) != 0) + _error->Errno("seteuid", "seteuid %u failed", old_euid); + if (setegid(old_egid) != 0) + _error->Errno("setegid", "setegid %u failed", old_egid); + + return tmp; +} + /*}}}*/ +FileFd* GetTempFile(std::string const &Prefix, bool ImmediateUnlink, FileFd * const TmpFd) /*{{{*/ +{ + return GetTempFile(Prefix, ImmediateUnlink, TmpFd, false); +} +FileFd* GetTempFile(std::string const &Prefix, bool ImmediateUnlink, FileFd * const TmpFd, bool Buffered) +{ + std::string fn; + std::string const tempdir = GetTempDir(); + int fd = -1; +#ifdef O_TMPFILE + if (ImmediateUnlink) + fd = open(tempdir.c_str(), O_RDWR|O_TMPFILE|O_EXCL|O_CLOEXEC, 0600); + if (fd < 0) +#endif + { + auto const suffix = Prefix.find(".XXXXXX."); + std::vector<char> buffer(tempdir.length() + 1 + Prefix.length() + (suffix == std::string::npos ? 7 : 0) + 1, '\0'); + if (suffix != std::string::npos) + { + if (snprintf(buffer.data(), buffer.size(), "%s/%s", tempdir.c_str(), Prefix.c_str()) > 0) + { + ssize_t const suffixlen = (buffer.size() - 1) - (tempdir.length() + 1 + suffix + 7); + if (likely(suffixlen > 0)) + fd = mkstemps(buffer.data(), suffixlen); + } + } + else + { + if (snprintf(buffer.data(), buffer.size(), "%s/%s.XXXXXX", tempdir.c_str(), Prefix.c_str()) > 0) + fd = mkstemp(buffer.data()); + } + fn.assign(buffer.data(), buffer.size() - 1); + if (ImmediateUnlink && fd != -1) + unlink(fn.c_str()); + } + if (fd < 0) + { + _error->Errno("GetTempFile",_("Unable to mkstemp %s"), fn.c_str()); + return nullptr; + } + FileFd * const Fd = TmpFd == nullptr ? new FileFd() : TmpFd; + if (not Fd->OpenDescriptor(fd, FileFd::ReadWrite | (Buffered ? FileFd::BufferedWrite : 0), FileFd::None, true)) + { + _error->Errno("GetTempFile",_("Unable to write to %s"),fn.c_str()); + if (TmpFd == nullptr) + delete Fd; + return nullptr; + } + if (not ImmediateUnlink) + Fd->SetFileName(fn); + return Fd; +} + /*}}}*/ +bool Rename(std::string From, std::string To) /*{{{*/ +{ + if (rename(From.c_str(),To.c_str()) != 0) + { + _error->Error(_("rename failed, %s (%s -> %s)."),strerror(errno), + From.c_str(),To.c_str()); + return false; + } + return true; +} + /*}}}*/ +bool Popen(const char *Args[], FileFd &Fd, pid_t &Child, FileFd::OpenMode Mode, bool CaptureStderr, bool Sandbox) /*{{{*/ +{ + int fd; + if (Mode != FileFd::ReadOnly && Mode != FileFd::WriteOnly) + return _error->Error("Popen supports ReadOnly (x)or WriteOnly mode only"); + + int Pipe[2] = {-1, -1}; + if(pipe(Pipe) != 0) + return _error->Errno("pipe", _("Failed to create subprocess IPC")); + + std::set<int> keep_fds; + keep_fds.insert(Pipe[0]); + keep_fds.insert(Pipe[1]); + Child = ExecFork(keep_fds); + if(Child < 0) + return _error->Errno("fork", "Failed to fork"); + if(Child == 0) + { + if (Sandbox && (getuid() == 0 || geteuid() == 0) && !DropPrivileges()) + { + _error->DumpErrors(); + _exit(1); + } + if(Mode == FileFd::ReadOnly) + { + close(Pipe[0]); + fd = Pipe[1]; + } + else if(Mode == FileFd::WriteOnly) + { + close(Pipe[1]); + fd = Pipe[0]; + } + + if(Mode == FileFd::ReadOnly) + { + dup2(fd, 1); + if (CaptureStderr == true) + dup2(fd, 2); + } else if(Mode == FileFd::WriteOnly) + dup2(fd, 0); + + execv(Args[0], (char**)Args); + _exit(100); + } + if(Mode == FileFd::ReadOnly) + { + close(Pipe[1]); + fd = Pipe[0]; + } + else if(Mode == FileFd::WriteOnly) + { + close(Pipe[0]); + fd = Pipe[1]; + } + else + return _error->Error("Popen supports ReadOnly (x)or WriteOnly mode only"); + Fd.OpenDescriptor(fd, Mode, FileFd::None, true); + + return true; +} + /*}}}*/ +bool DropPrivileges() /*{{{*/ +{ + if(_config->FindB("Debug::NoDropPrivs", false) == true) + return true; + +#if __gnu_linux__ +#if defined(PR_SET_NO_NEW_PRIVS) && ( PR_SET_NO_NEW_PRIVS != 38 ) +#error "PR_SET_NO_NEW_PRIVS is defined, but with a different value than expected!" +#endif + // see prctl(2), needs linux3.5 at runtime - magic constant to avoid it at buildtime + int ret = prctl(38, 1, 0, 0, 0); + // ignore EINVAL - kernel is too old to understand the option + if(ret < 0 && errno != EINVAL) + _error->Warning("PR_SET_NO_NEW_PRIVS failed with %i", ret); +#endif + + // empty setting disables privilege dropping - this also ensures + // backward compatibility, see bug #764506 + const std::string toUser = _config->Find("APT::Sandbox::User"); + if (toUser.empty() || toUser == "root") + return true; + + // a lot can go wrong trying to drop privileges completely, + // so ideally we would like to verify that we have done it – + // but the verify asks for too much in case of fakeroot (and alike) + // [Specific checks can be overridden with dedicated options] + bool const VerifySandboxing = _config->FindB("APT::Sandbox::Verify", false); + + // uid will be 0 in the end, but gid might be different anyway + uid_t const old_uid = getuid(); + gid_t const old_gid = getgid(); + + if (old_uid != 0) + return true; + + struct passwd *pw = getpwnam(toUser.c_str()); + if (pw == NULL) + return _error->Error("No user %s, can not drop rights", toUser.c_str()); + + // Do not change the order here, it might break things + // Get rid of all our supplementary groups first + if (setgroups(1, &pw->pw_gid)) + return _error->Errno("setgroups", "Failed to setgroups"); + + // Now change the group ids to the new user +#ifdef HAVE_SETRESGID + if (setresgid(pw->pw_gid, pw->pw_gid, pw->pw_gid) != 0) + return _error->Errno("setresgid", "Failed to set new group ids"); +#else + if (setegid(pw->pw_gid) != 0) + return _error->Errno("setegid", "Failed to setegid"); + + if (setgid(pw->pw_gid) != 0) + return _error->Errno("setgid", "Failed to setgid"); +#endif + + // Change the user ids to the new user +#ifdef HAVE_SETRESUID + if (setresuid(pw->pw_uid, pw->pw_uid, pw->pw_uid) != 0) + return _error->Errno("setresuid", "Failed to set new user ids"); +#else + if (setuid(pw->pw_uid) != 0) + return _error->Errno("setuid", "Failed to setuid"); + if (seteuid(pw->pw_uid) != 0) + return _error->Errno("seteuid", "Failed to seteuid"); +#endif + + // disabled by default as fakeroot doesn't implement getgroups currently (#806521) + if (VerifySandboxing == true || _config->FindB("APT::Sandbox::Verify::Groups", false) == true) + { + // Verify that the user isn't still in any supplementary groups + long const ngroups_max = sysconf(_SC_NGROUPS_MAX); + std::unique_ptr<gid_t[]> gidlist(new gid_t[ngroups_max]); + if (unlikely(gidlist == NULL)) + return _error->Error("Allocation of a list of size %lu for getgroups failed", ngroups_max); + ssize_t gidlist_nr; + if ((gidlist_nr = getgroups(ngroups_max, gidlist.get())) < 0) + return _error->Errno("getgroups", "Could not get new groups (%lu)", ngroups_max); + for (ssize_t i = 0; i < gidlist_nr; ++i) + if (gidlist[i] != pw->pw_gid) + return _error->Error("Could not switch group, user %s is still in group %d", toUser.c_str(), gidlist[i]); + } + + // enabled by default as all fakeroot-lookalikes should fake that accordingly + if (VerifySandboxing == true || _config->FindB("APT::Sandbox::Verify::IDs", true) == true) + { + // Verify that gid, egid, uid, and euid changed + if (getgid() != pw->pw_gid) + return _error->Error("Could not switch group"); + if (getegid() != pw->pw_gid) + return _error->Error("Could not switch effective group"); + if (getuid() != pw->pw_uid) + return _error->Error("Could not switch user"); + if (geteuid() != pw->pw_uid) + return _error->Error("Could not switch effective user"); + +#ifdef HAVE_GETRESUID + // verify that the saved set-user-id was changed as well + uid_t ruid = 0; + uid_t euid = 0; + uid_t suid = 0; + if (getresuid(&ruid, &euid, &suid)) + return _error->Errno("getresuid", "Could not get saved set-user-ID"); + if (suid != pw->pw_uid) + return _error->Error("Could not switch saved set-user-ID"); +#endif + +#ifdef HAVE_GETRESGID + // verify that the saved set-group-id was changed as well + gid_t rgid = 0; + gid_t egid = 0; + gid_t sgid = 0; + if (getresgid(&rgid, &egid, &sgid)) + return _error->Errno("getresuid", "Could not get saved set-group-ID"); + if (sgid != pw->pw_gid) + return _error->Error("Could not switch saved set-group-ID"); +#endif + } + + // disabled as fakeroot doesn't forbid (by design) (re)gaining root from unprivileged + if (VerifySandboxing == true || _config->FindB("APT::Sandbox::Verify::Regain", false) == true) + { + // Check that uid and gid changes do not work anymore + if (pw->pw_gid != old_gid && (setgid(old_gid) != -1 || setegid(old_gid) != -1)) + return _error->Error("Could restore a gid to root, privilege dropping did not work"); + + if (pw->pw_uid != old_uid && (setuid(old_uid) != -1 || seteuid(old_uid) != -1)) + return _error->Error("Could restore a uid to root, privilege dropping did not work"); + } + + if (_config->FindB("APT::Sandbox::ResetEnvironment", true)) + { + setenv("HOME", pw->pw_dir, 1); + setenv("USER", pw->pw_name, 1); + setenv("USERNAME", pw->pw_name, 1); + setenv("LOGNAME", pw->pw_name, 1); + auto const shell = flNotDir(pw->pw_shell); + if (shell == "false" || shell == "nologin") + setenv("SHELL", "/bin/sh", 1); + else + setenv("SHELL", pw->pw_shell, 1); + auto const apt_setenv_tmp = [](char const * const env) { + auto const tmpdir = getenv(env); + if (tmpdir != nullptr) + { + auto const ourtmpdir = GetTempDirEnv(env); + if (ourtmpdir != tmpdir) + setenv(env, ourtmpdir.c_str(), 1); + } + }; + apt_setenv_tmp("TMPDIR"); + apt_setenv_tmp("TEMPDIR"); + apt_setenv_tmp("TMP"); + apt_setenv_tmp("TEMP"); + } + + return true; +} + /*}}}*/ +bool OpenConfigurationFileFd(std::string const &File, FileFd &Fd) /*{{{*/ +{ + int const fd = open(File.c_str(), O_RDONLY | O_CLOEXEC | O_NOCTTY); + if (fd == -1) + return _error->WarningE("open", _("Unable to read %s"), File.c_str()); + APT::Configuration::Compressor none(".", "", "", nullptr, nullptr, 0); + if (Fd.OpenDescriptor(fd, FileFd::ReadOnly, none, true) == false) + return false; + Fd.SetFileName(File); + return true; +} + /*}}}*/ +int Inhibit(const char *what, const char *who, const char *why, const char *mode) /*{{{*/ +{ +#ifdef HAVE_SYSTEMD + sd_bus_error error = SD_BUS_ERROR_NULL; + sd_bus_message *m = NULL; + sd_bus *bus = NULL; + int fd; + int r; + + r = sd_bus_open_system(&bus); + if (r < 0) + goto out; + + r = sd_bus_call_method(bus, + "org.freedesktop.login1", + "/org/freedesktop/login1", + "org.freedesktop.login1.Manager", + "Inhibit", + &error, + &m, + "ssss", + what, + who, + why, + mode); + if (r < 0) + goto out; + + r = sd_bus_message_read(m, "h", &fd); + if (r < 0) + goto out; + + // We received a file descriptor, return it - systemd will close the read fd + // on free, so let's duplicate it here. + r = dup(fd); +out: + sd_bus_error_free(&error); + sd_bus_message_unref(m); + sd_bus_unref(bus); + return r; +#else + return -ENOTSUP; +#endif +} + /*}}}*/ diff --git a/apt-pkg/contrib/fileutl.h b/apt-pkg/contrib/fileutl.h new file mode 100644 index 0000000..970b118 --- /dev/null +++ b/apt-pkg/contrib/fileutl.h @@ -0,0 +1,287 @@ +// -*- mode: cpp; mode: fold -*- +// SPDX-License-Identifier: GPL-2.0+ +// Description /*{{{*/ +/* ###################################################################### + + File Utilities + + CopyFile - Buffered copy of a single file + GetLock - dpkg compatible lock file manipulation (fcntl) + FileExists - Returns true if the file exists + SafeGetCWD - Returns the CWD in a string with overrun protection + + The file class is a handy abstraction for various functions+classes + that need to accept filenames. + + This file had this historic note, but now includes further changes + under the GPL-2.0+: + + This source is placed in the Public Domain, do with it what you will + It was originally written by Jason Gunthorpe. + + ##################################################################### */ + /*}}}*/ +#ifndef PKGLIB_FILEUTL_H +#define PKGLIB_FILEUTL_H + +#include <apt-pkg/aptconfiguration.h> +#include <apt-pkg/macros.h> + +#include <set> +#include <string> +#include <vector> +#include <sys/stat.h> +#include <time.h> + +/* Define this for python-apt */ +#define APT_HAS_GZIP 1 + +class FileFdPrivate; +class APT_PUBLIC FileFd +{ + friend class FileFdPrivate; + friend class GzipFileFdPrivate; + friend class Bz2FileFdPrivate; + friend class LzmaFileFdPrivate; + friend class Lz4FileFdPrivate; + friend class ZstdFileFdPrivate; + friend class DirectFileFdPrivate; + friend class PipedFileFdPrivate; + protected: + int iFd; + + enum LocalFlags {AutoClose = (1<<0),Fail = (1<<1),DelOnFail = (1<<2), + HitEof = (1<<3), Replace = (1<<4), Compressed = (1<<5) }; + unsigned long Flags; + std::string FileName; + std::string TemporaryFileName; + + public: + enum OpenMode { + ReadOnly = (1 << 0), + WriteOnly = (1 << 1), + ReadWrite = ReadOnly | WriteOnly, + + Create = (1 << 2), + Exclusive = (1 << 3), + Atomic = Exclusive | (1 << 4), + Empty = (1 << 5), + BufferedWrite = (1 << 6), + + WriteEmpty = ReadWrite | Create | Empty, + WriteExists = ReadWrite, + WriteAny = ReadWrite | Create, + WriteTemp = ReadWrite | Create | Exclusive, + ReadOnlyGzip, + WriteAtomic = ReadWrite | Create | Atomic + }; + enum CompressMode + { + Auto = 'A', + None = 'N', + Extension = 'E', + Gzip = 'G', + Bzip2 = 'B', + Lzma = 'L', + Xz = 'X', + Lz4 = '4', + Zstd = 'Z' + }; + + inline bool Read(void *To,unsigned long long Size,bool AllowEof) + { + unsigned long long Jnk; + if (AllowEof) + return Read(To,Size,&Jnk); + return Read(To,Size); + } + bool Read(void *To,unsigned long long Size,unsigned long long *Actual = 0); + bool static Read(int const Fd, void *To, unsigned long long Size, unsigned long long * const Actual = 0); + /** read a complete line or until buffer is full + * + * The buffer will always be \\0 terminated, so at most Size-1 characters are read. + * If the buffer holds a complete line the last character (before \\0) will be + * the newline character \\n otherwise the line was longer than the buffer. + * + * @param To buffer which will hold the line + * @param Size of the buffer to fill + * @param \b nullptr is returned in error cases, otherwise + * the parameter \b To now filled with the line. + */ + char* ReadLine(char *To, unsigned long long const Size); + /** read a complete line from the file + * + * Similar to std::getline() the string does \b not include + * the newline, but just the content of the line as the newline + * is not needed to distinguish cases as for the other #ReadLine method. + * + * @param To string which will hold the line + * @return \b true if successful, otherwise \b false + */ + bool ReadLine(std::string &To); + bool Flush(); + bool Write(const void *From,unsigned long long Size); + bool static Write(int Fd, const void *From, unsigned long long Size); + bool Seek(unsigned long long To); + bool Skip(unsigned long long To); + bool Truncate(unsigned long long To); + unsigned long long Tell(); + // the size of the file content (compressed files will be uncompressed first) + unsigned long long Size(); + // the size of the file itself + unsigned long long FileSize(); + time_t ModificationTime(); + + bool Open(std::string FileName,unsigned int const Mode,CompressMode Compress,unsigned long const AccessMode = 0666); + bool Open(std::string FileName,unsigned int const Mode,APT::Configuration::Compressor const &compressor,unsigned long const AccessMode = 0666); + inline bool Open(std::string const &FileName,unsigned int const Mode, unsigned long const AccessMode = 0666) { + return Open(FileName, Mode, None, AccessMode); + }; + bool OpenDescriptor(int Fd, unsigned int const Mode, CompressMode Compress, bool AutoClose=false); + bool OpenDescriptor(int Fd, unsigned int const Mode, APT::Configuration::Compressor const &compressor, bool AutoClose=false); + inline bool OpenDescriptor(int Fd, unsigned int const Mode, bool AutoClose=false) { + return OpenDescriptor(Fd, Mode, None, AutoClose); + }; + bool Close(); + bool Sync(); + + // Simple manipulators + inline int Fd() {return iFd;}; + inline void Fd(int fd) { OpenDescriptor(fd, ReadWrite);}; + + inline bool IsOpen() {return iFd >= 0;}; + inline bool Failed() {return (Flags & Fail) == Fail;}; + inline void EraseOnFailure() {Flags |= DelOnFail;}; + inline void OpFail() {Flags |= Fail;}; + inline bool Eof() {return (Flags & HitEof) == HitEof;}; + inline bool IsCompressed() {return (Flags & Compressed) == Compressed;}; + inline std::string &Name() {return FileName;}; + inline void SetFileName(std::string const &name) { FileName = name; }; + + FileFd(std::string FileName,unsigned int const Mode,unsigned long AccessMode = 0666); + FileFd(std::string FileName,unsigned int const Mode, CompressMode Compress, unsigned long AccessMode = 0666); + FileFd(); + FileFd(int const Fd, unsigned int const Mode = ReadWrite, CompressMode Compress = None); + FileFd(int const Fd, bool const AutoClose); + virtual ~FileFd(); + + private: + FileFdPrivate * d; + APT_HIDDEN FileFd(const FileFd &); + APT_HIDDEN FileFd & operator=(const FileFd &); + APT_HIDDEN bool OpenInternDescriptor(unsigned int const Mode, APT::Configuration::Compressor const &compressor); + + // private helpers to set Fail flag and call _error->Error + APT_HIDDEN bool FileFdErrno(const char* Function, const char* Description,...) APT_PRINTF(3) APT_COLD; + APT_HIDDEN bool FileFdError(const char* Description,...) APT_PRINTF(2) APT_COLD; +}; + +APT_PUBLIC bool RunScripts(const char *Cnf); +APT_PUBLIC bool CopyFile(FileFd &From,FileFd &To); +APT_PUBLIC bool RemoveFile(char const * const Function, std::string const &FileName); +APT_PUBLIC bool RemoveFileAt(char const * const Function, int const dirfd, std::string const &FileName); +APT_PUBLIC int GetLock(std::string File,bool Errors = true); +APT_PUBLIC bool FileExists(std::string File); +APT_PUBLIC bool RealFileExists(std::string File); +APT_PUBLIC bool DirectoryExists(std::string const &Path); +APT_PUBLIC bool CreateDirectory(std::string const &Parent, std::string const &Path); +APT_PUBLIC time_t GetModificationTime(std::string const &Path); +APT_PUBLIC bool Rename(std::string From, std::string To); + +APT_PUBLIC std::string GetTempDir(); +APT_PUBLIC std::string GetTempDir(std::string const &User); +APT_PUBLIC FileFd* GetTempFile(std::string const &Prefix = "", + bool ImmediateUnlink = true, + FileFd * const TmpFd = NULL); + +// FIXME: GetTempFile should always return a buffered file +APT_HIDDEN FileFd* GetTempFile(std::string const &Prefix, + bool ImmediateUnlink , + FileFd * const TmpFd, + bool Buffered); + +/** \brief Ensure the existence of the given Path + * + * \param Parent directory of the Path directory - a trailing + * /apt/ will be removed before CreateDirectory call. + * \param Path which should exist after (successful) call + */ +APT_PUBLIC bool CreateAPTDirectoryIfNeeded(std::string const &Parent, std::string const &Path); + +APT_PUBLIC std::vector<std::string> GetListOfFilesInDir(std::string const &Dir, std::string const &Ext, + bool const &SortList, bool const &AllowNoExt=false); +APT_PUBLIC std::vector<std::string> GetListOfFilesInDir(std::string const &Dir, std::vector<std::string> const &Ext, + bool const &SortList); +APT_PUBLIC std::vector<std::string> GetListOfFilesInDir(std::string const &Dir, bool SortList); +APT_PUBLIC std::string SafeGetCWD(); +APT_PUBLIC void SetCloseExec(int Fd,bool Close); +APT_PUBLIC void SetNonBlock(int Fd,bool Block); +APT_PUBLIC bool WaitFd(int Fd,bool write = false,unsigned long timeout = 0); +APT_PUBLIC pid_t ExecFork(); +APT_PUBLIC pid_t ExecFork(std::set<int> keep_fds); +APT_PUBLIC void MergeKeepFdsFromConfiguration(std::set<int> &keep_fds); +APT_PUBLIC bool ExecWait(pid_t Pid,const char *Name,bool Reap = false); + +// check if the given file starts with a PGP cleartext signature +APT_PUBLIC bool StartsWithGPGClearTextSignature(std::string const &FileName); + +/** change file attributes to requested known good values + * + * The method skips the user:group setting if not root. + * + * @param requester is printed as functionname in error cases + * @param file is the file to be modified + * @param user is the (new) owner of the file, e.g. _apt + * @param group is the (new) group owning the file, e.g. root + * @param mode is the access mode of the file, e.g. 0644 + */ +APT_PUBLIC bool ChangeOwnerAndPermissionOfFile(char const * const requester, char const * const file, char const * const user, char const * const group, mode_t const mode); + +/** + * \brief Drop privileges + * + * Drop the privileges to the user _apt (or the one specified in + * APT::Sandbox::User). This does not set the supplementary group + * ids up correctly, it only uses the default group. Also prevent + * the process from gaining any new privileges afterwards, at least + * on Linux. + * + * \return true on success, false on failure with _error set + */ +APT_PUBLIC bool DropPrivileges(); + +// File string manipulators +APT_PUBLIC std::string flNotDir(std::string File); +APT_PUBLIC std::string flNotFile(std::string File); +APT_PUBLIC std::string flNoLink(std::string File); +APT_PUBLIC std::string flExtension(std::string File); +APT_PUBLIC std::string flCombine(std::string Dir,std::string File); + +/** \brief Takes a file path and returns the absolute path + */ +APT_PUBLIC std::string flAbsPath(std::string File); +/** \brief removes superfluous /./ and // from path */ +APT_HIDDEN std::string flNormalize(std::string file); + +// simple c++ glob +APT_PUBLIC std::vector<std::string> Glob(std::string const &pattern, int flags=0); + +/** \brief Popen() implementation that execv() instead of using a shell + * + * \param Args the execv style command to run + * \param FileFd is a reference to the FileFd to use for input or output + * \param Child a reference to the integer that stores the child pid + * Note that you must call ExecWait() or similar to cleanup + * \param Mode is either FileFd::ReadOnly or FileFd::WriteOnly + * \param CaptureStderr True if we should capture stderr in addition to stdout. + * (default: True). + * \param Sandbox True if this should run sandboxed + * \return true on success, false on failure with _error set + */ +APT_PUBLIC bool Popen(const char *Args[], FileFd &Fd, pid_t &Child, FileFd::OpenMode Mode, bool CaptureStderr = true, bool Sandbox = false); + +APT_HIDDEN bool OpenConfigurationFileFd(std::string const &File, FileFd &Fd); + +APT_HIDDEN int Inhibit(const char *what, const char *who, const char *why, const char *mode); + +#endif diff --git a/apt-pkg/contrib/gpgv.cc b/apt-pkg/contrib/gpgv.cc new file mode 100644 index 0000000..3368ece --- /dev/null +++ b/apt-pkg/contrib/gpgv.cc @@ -0,0 +1,568 @@ +// -*- mode: cpp; mode: fold -*- +// Include Files /*{{{*/ +#include <config.h> + +#include <apt-pkg/configuration.h> +#include <apt-pkg/error.h> +#include <apt-pkg/fileutl.h> +#include <apt-pkg/gpgv.h> +#include <apt-pkg/strutl.h> + +#include <errno.h> +#include <fcntl.h> +#include <stddef.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <sys/wait.h> +#include <unistd.h> + +#include <algorithm> +#include <fstream> +#include <iostream> +#include <memory> +#include <sstream> +#include <string> +#include <vector> + +#include <apti18n.h> + /*}}}*/ + +// syntactic sugar to wrap a raw pointer with a custom deleter in a std::unique_ptr +static std::unique_ptr<char, decltype(&free)> make_unique_char(void *const str = nullptr) +{ + return {static_cast<char *>(str), &free}; +} +static std::unique_ptr<FILE, decltype(&fclose)> make_unique_FILE(std::string const &filename, char const *const mode) +{ + return {fopen(filename.c_str(), mode), &fclose}; +} + +class LineBuffer /*{{{*/ +{ + char *buffer = nullptr; + size_t buffer_size = 0; + int line_length = 0; + // a "normal" find_last_not_of returns npos if not found + int find_last_not_of_length(APT::StringView const bad) const + { + for (int result = line_length - 1; result >= 0; --result) + if (bad.find(buffer[result]) == APT::StringView::npos) + return result + 1; + return 0; + } + + public: + bool empty() const noexcept { return view().empty(); } + APT::StringView view() const noexcept { return {buffer, static_cast<size_t>(line_length)}; } + bool starts_with(APT::StringView const start) const { return view().substr(0, start.size()) == start; } + + bool writeTo(FileFd *const to, size_t offset = 0) const + { + if (to == nullptr) + return true; + return to->Write(buffer + offset, line_length - offset); + } + bool writeLineTo(FileFd *const to) const + { + if (to == nullptr) + return true; + buffer[line_length] = '\n'; + bool const result = to->Write(buffer, line_length + 1); + buffer[line_length] = '\0'; + return result; + } + bool writeNewLineIf(FileFd *const to, bool const condition) const + { + if (not condition || to == nullptr) + return true; + return to->Write("\n", 1); + } + + bool readFrom(FILE *stream, std::string const &InFile, bool acceptEoF = false) + { + errno = 0; + line_length = getline(&buffer, &buffer_size, stream); + if (errno != 0) + return _error->Errno("getline", "Could not read from %s", InFile.c_str()); + if (line_length == -1) + { + if (acceptEoF) + return false; + return _error->Error("Splitting of clearsigned file %s failed as it doesn't contain all expected parts", InFile.c_str()); + } + // a) remove newline characters, so we can work consistently with lines + line_length = find_last_not_of_length("\n\r"); + // b) remove trailing whitespaces as defined by rfc4880 §7.1 + line_length = find_last_not_of_length(" \t"); + buffer[line_length] = '\0'; + return true; + } + + ~LineBuffer() { free(buffer); } +}; +static bool operator==(LineBuffer const &buf, APT::StringView const exp) noexcept +{ + return buf.view() == exp; +} +static bool operator!=(LineBuffer const &buf, APT::StringView const exp) noexcept +{ + return buf.view() != exp; +} + /*}}}*/ +// ExecGPGV - returns the command needed for verify /*{{{*/ +// --------------------------------------------------------------------- +/* Generating the commandline for calling gpg is somehow complicated as + we need to add multiple keyrings and user supplied options. + Also, as gpg has no options to enforce a certain reduced style of + clear-signed files (=the complete content of the file is signed and + the content isn't encoded) we do a divide and conquer approach here + and split up the clear-signed file in message and signature for gpg. + And as a cherry on the cake, we use our apt-key wrapper to do part + of the lifting in regards to merging keyrings. Fun for the whole family. +*/ +static void APT_PRINTF(4) apt_error(std::ostream &outterm, int const statusfd, int fd[2], const char *format, ...) +{ + std::ostringstream outstr; + std::ostream &out = (statusfd == -1) ? outterm : outstr; + va_list args; + ssize_t size = 400; + while (true) { + bool ret; + va_start(args,format); + ret = iovprintf(out, format, args, size); + va_end(args); + if (ret) + break; + } + if (statusfd != -1) + { + auto const errtag = "[APTKEY:] ERROR "; + outstr << '\n'; + auto const errtext = outstr.str(); + if (not FileFd::Write(fd[1], errtag, strlen(errtag)) || + not FileFd::Write(fd[1], errtext.data(), errtext.size())) + outterm << errtext << std::flush; + } +} +void ExecGPGV(std::string const &File, std::string const &FileGPG, + int const &statusfd, int fd[2], std::string const &key) +{ + #define EINTERNAL 111 + std::string const aptkey = _config->Find("Dir::Bin::apt-key", CMAKE_INSTALL_FULL_BINDIR "/apt-key"); + + bool const Debug = _config->FindB("Debug::Acquire::gpgv", false); + struct exiter { + std::vector<const char *> files; + void operator ()(int code) APT_NORETURN { + std::for_each(files.begin(), files.end(), unlink); + exit(code); + } + } local_exit; + + + std::vector<const char *> Args; + Args.reserve(10); + + Args.push_back(aptkey.c_str()); + Args.push_back("--quiet"); + Args.push_back("--readonly"); + auto const keysFileFpr = VectorizeString(key, ','); + for (auto const &k: keysFileFpr) + { + if (unlikely(k.empty())) + continue; + if (k[0] == '/') + { + Args.push_back("--keyring"); + Args.push_back(k.c_str()); + } + else + { + Args.push_back("--keyid"); + Args.push_back(k.c_str()); + } + } + Args.push_back("verify"); + + char statusfdstr[10]; + if (statusfd != -1) + { + Args.push_back("--status-fd"); + snprintf(statusfdstr, sizeof(statusfdstr), "%i", statusfd); + Args.push_back(statusfdstr); + } + + Configuration::Item const *Opts; + Opts = _config->Tree("Acquire::gpgv::Options"); + if (Opts != 0) + { + Opts = Opts->Child; + for (; Opts != 0; Opts = Opts->Next) + { + if (Opts->Value.empty()) + continue; + Args.push_back(Opts->Value.c_str()); + } + } + + enum { DETACHED, CLEARSIGNED } releaseSignature = (FileGPG != File) ? DETACHED : CLEARSIGNED; + auto sig = make_unique_char(); + auto data = make_unique_char(); + auto conf = make_unique_char(); + + // Dump the configuration so apt-key picks up the correct Dir values + { + { + std::string tmpfile; + strprintf(tmpfile, "%s/apt.conf.XXXXXX", GetTempDir().c_str()); + conf.reset(strdup(tmpfile.c_str())); + } + if (conf == nullptr) { + apt_error(std::cerr, statusfd, fd, "Couldn't create tempfile names for passing config to apt-key"); + local_exit(EINTERNAL); + } + int confFd = mkstemp(conf.get()); + if (confFd == -1) { + apt_error(std::cerr, statusfd, fd, "Couldn't create temporary file %s for passing config to apt-key", conf.get()); + local_exit(EINTERNAL); + } + local_exit.files.push_back(conf.get()); + + std::ofstream confStream(conf.get()); + close(confFd); + _config->Dump(confStream); + confStream.close(); + setenv("APT_CONFIG", conf.get(), 1); + } + + // Tell apt-key not to emit warnings + setenv("APT_KEY_DONT_WARN_ON_DANGEROUS_USAGE", "1", 1); + + if (releaseSignature == DETACHED) + { + auto detached = make_unique_FILE(FileGPG, "r"); + if (detached.get() == nullptr) + { + apt_error(std::cerr, statusfd, fd, "Detached signature file '%s' could not be opened", FileGPG.c_str()); + local_exit(EINTERNAL); + } + LineBuffer buf; + bool open_signature = false; + bool found_badcontent = false; + size_t found_signatures = 0; + while (buf.readFrom(detached.get(), FileGPG, true)) + { + if (open_signature) + { + if (buf == "-----END PGP SIGNATURE-----") + open_signature = false; + else if (buf.starts_with("-")) + { + // the used Radix-64 is not using dash for any value, so a valid line can't + // start with one. Header keys could, but no existent one does and seems unlikely. + // Instead it smells a lot like a header the parser didn't recognize. + apt_error(std::cerr, statusfd, fd, "Detached signature file '%s' contains unexpected line starting with a dash", FileGPG.c_str()); + local_exit(112); + } + } + else //if (not open_signature) + { + if (buf == "-----BEGIN PGP SIGNATURE-----") + { + open_signature = true; + ++found_signatures; + if (found_badcontent) + break; + } + else + { + found_badcontent = true; + if (found_signatures != 0) + break; + } + } + } + if (found_signatures == 0 && statusfd != -1) + { + auto const errtag = "[GNUPG:] NODATA\n"; + FileFd::Write(fd[1], errtag, strlen(errtag)); + // guess if this is a binary signature, we never officially supported them, + // but silently accepted them via passing them unchecked to gpgv + if (found_badcontent) + { + rewind(detached.get()); + auto ptag = fgetc(detached.get()); + // §4.2 says that the first bit is always set and gpg seems to generate + // only old format which is indicated by the second bit not set + if (ptag != EOF && (ptag & 0x80) != 0 && (ptag & 0x40) == 0) + { + apt_error(std::cerr, statusfd, fd, "Detached signature file '%s' is in unsupported binary format", FileGPG.c_str()); + local_exit(112); + } + } + // This is not an attack attempt but a file even gpgv would complain about + // likely the result of a paywall which is covered by the gpgv method + local_exit(113); + } + else if (found_badcontent) + { + apt_error(std::cerr, statusfd, fd, "Detached signature file '%s' contains lines not belonging to a signature", FileGPG.c_str()); + local_exit(112); + } + if (open_signature) + { + apt_error(std::cerr, statusfd, fd, "Detached signature file '%s' contains unclosed signatures", FileGPG.c_str()); + local_exit(112); + } + + Args.push_back(FileGPG.c_str()); + Args.push_back(File.c_str()); + } + else // clear-signed file + { + FileFd signature; + if (GetTempFile("apt.sig", false, &signature) == nullptr) + local_exit(EINTERNAL); + sig.reset(strdup(signature.Name().c_str())); + local_exit.files.push_back(sig.get()); + FileFd message; + if (GetTempFile("apt.data", false, &message) == nullptr) + local_exit(EINTERNAL); + data.reset(strdup(message.Name().c_str())); + local_exit.files.push_back(data.get()); + + if (signature.Failed() || message.Failed() || + not SplitClearSignedFile(File, &message, nullptr, &signature)) + { + apt_error(std::cerr, statusfd, fd, "Splitting up %s into data and signature failed", File.c_str()); + local_exit(112); + } + Args.push_back(sig.get()); + Args.push_back(data.get()); + } + + Args.push_back(NULL); + + if (Debug) + { + std::clog << "Preparing to exec: "; + for (std::vector<const char *>::const_iterator a = Args.begin(); *a != NULL; ++a) + std::clog << " " << *a; + std::clog << std::endl; + } + + if (statusfd != -1) + { + int const nullfd = open("/dev/null", O_WRONLY); + close(fd[0]); + // Redirect output to /dev/null; we read from the status fd + if (statusfd != STDOUT_FILENO) + dup2(nullfd, STDOUT_FILENO); + if (statusfd != STDERR_FILENO) + dup2(nullfd, STDERR_FILENO); + // Redirect the pipe to the status fd (3) + dup2(fd[1], statusfd); + + putenv((char *)"LANG="); + putenv((char *)"LC_ALL="); + putenv((char *)"LC_MESSAGES="); + } + + + // We have created tempfiles we have to clean up + // and we do an additional check, so fork yet another time … + pid_t pid = ExecFork(); + if(pid < 0) { + apt_error(std::cerr, statusfd, fd, "Fork failed for %s to check %s", Args[0], File.c_str()); + local_exit(EINTERNAL); + } + if(pid == 0) + { + if (statusfd != -1) + dup2(fd[1], statusfd); + execvp(Args[0], (char **) &Args[0]); + apt_error(std::cerr, statusfd, fd, "Couldn't execute %s to check %s", Args[0], File.c_str()); + local_exit(EINTERNAL); + } + + // Wait and collect the error code - taken from WaitPid as we need the exact Status + int Status; + while (waitpid(pid,&Status,0) != pid) + { + if (errno == EINTR) + continue; + apt_error(std::cerr, statusfd, fd, _("Waited for %s but it wasn't there"), "apt-key"); + local_exit(EINTERNAL); + } + + // check if it exit'ed normally … + if (not WIFEXITED(Status)) + { + apt_error(std::cerr, statusfd, fd, _("Sub-process %s exited unexpectedly"), "apt-key"); + local_exit(EINTERNAL); + } + + // … and with a good exit code + if (WEXITSTATUS(Status) != 0) + { + // we forward the statuscode, so don't generate a message on the fd in this case + apt_error(std::cerr, -1, fd, _("Sub-process %s returned an error code (%u)"), "apt-key", WEXITSTATUS(Status)); + local_exit(WEXITSTATUS(Status)); + } + + // everything fine + local_exit(0); +} + /*}}}*/ +// SplitClearSignedFile - split message into data/signature /*{{{*/ +bool SplitClearSignedFile(std::string const &InFile, FileFd * const ContentFile, + std::vector<std::string> * const ContentHeader, FileFd * const SignatureFile) +{ + auto in = make_unique_FILE(InFile, "r"); + if (in.get() == nullptr) + return _error->Errno("fopen", "can not open %s", InFile.c_str()); + + struct ScopedErrors + { + ScopedErrors() { _error->PushToStack(); } + ~ScopedErrors() { _error->MergeWithStack(); } + } scoped; + LineBuffer buf; + + // start of the message + if (not buf.readFrom(in.get(), InFile)) + return false; // empty or read error + if (buf != "-----BEGIN PGP SIGNED MESSAGE-----") + { + // this might be an unsigned file we don't want to report errors for, + // but still finish unsuccessful none the less. + while (buf.readFrom(in.get(), InFile, true)) + if (buf == "-----BEGIN PGP SIGNED MESSAGE-----") + return _error->Error("Clearsigned file '%s' does not start with a signed message block.", InFile.c_str()); + + return false; + } + + // save "Hash" Armor Headers + while (true) + { + if (not buf.readFrom(in.get(), InFile)) + return false; + if (buf.empty()) + break; // empty line ends the Armor Headers + if (buf.starts_with("-")) + // § 6.2 says unknown keys should be reported to the user. We don't go that far, + // but we assume that there will never be a header key starting with a dash + return _error->Error("Clearsigned file '%s' contains unexpected line starting with a dash (%s)", InFile.c_str(), "armor"); + if (ContentHeader != nullptr && buf.starts_with("Hash: ")) + ContentHeader->push_back(buf.view().to_string()); + } + + // the message itself + bool first_line = true; + while (true) + { + if (not buf.readFrom(in.get(), InFile)) + return false; + + if (buf.starts_with("-")) + { + if (buf == "-----BEGIN PGP SIGNATURE-----") + { + if (not buf.writeLineTo(SignatureFile)) + return false; + break; + } + else if (buf.starts_with("- ")) + { + // we don't have any fields which need to be dash-escaped, + // but implementations are free to escape all lines … + if (not buf.writeNewLineIf(ContentFile, not first_line) || not buf.writeTo(ContentFile, 2)) + return false; + } + else + // § 7.1 says a client should warn, but we don't really work with files which + // should contain lines starting with a dash, so it is a lot more likely that + // this is an attempt to trick our parser vs. gpgv parser into ignoring a header + return _error->Error("Clearsigned file '%s' contains unexpected line starting with a dash (%s)", InFile.c_str(), "msg"); + } + else if (not buf.writeNewLineIf(ContentFile, not first_line) || not buf.writeTo(ContentFile)) + return false; + first_line = false; + } + + // collect all signatures + bool open_signature = true; + while (true) + { + if (not buf.readFrom(in.get(), InFile, true)) + break; + + if (open_signature) + { + if (buf == "-----END PGP SIGNATURE-----") + open_signature = false; + else if (buf.starts_with("-")) + // the used Radix-64 is not using dash for any value, so a valid line can't + // start with one. Header keys could, but no existent one does and seems unlikely. + // Instead it smells a lot like a header the parser didn't recognize. + return _error->Error("Clearsigned file '%s' contains unexpected line starting with a dash (%s)", InFile.c_str(), "sig"); + } + else //if (not open_signature) + { + if (buf == "-----BEGIN PGP SIGNATURE-----") + open_signature = true; + else + return _error->Error("Clearsigned file '%s' contains unsigned lines.", InFile.c_str()); + } + + if (not buf.writeLineTo(SignatureFile)) + return false; + } + if (open_signature) + return _error->Error("Signature in file %s wasn't closed", InFile.c_str()); + + // Flush the files + if (SignatureFile != nullptr) + SignatureFile->Flush(); + if (ContentFile != nullptr) + ContentFile->Flush(); + + // Catch-all for "unhandled" read/sync errors + if (_error->PendingError()) + return false; + return true; +} + /*}}}*/ +bool OpenMaybeClearSignedFile(std::string const &ClearSignedFileName, FileFd &MessageFile) /*{{{*/ +{ + // Buffered file + if (GetTempFile("clearsigned.message", true, &MessageFile, true) == nullptr) + return false; + if (MessageFile.Failed()) + return _error->Error("Couldn't open temporary file to work with %s", ClearSignedFileName.c_str()); + + _error->PushToStack(); + bool const splitDone = SplitClearSignedFile(ClearSignedFileName, &MessageFile, NULL, NULL); + bool const errorDone = _error->PendingError(); + _error->MergeWithStack(); + if (not splitDone) + { + MessageFile.Close(); + + if (errorDone) + return false; + + // we deal with an unsigned file + MessageFile.Open(ClearSignedFileName, FileFd::ReadOnly); + } + else // clear-signed + { + if (not MessageFile.Seek(0)) + return _error->Errno("lseek", "Unable to seek back in message for file %s", ClearSignedFileName.c_str()); + } + + return not MessageFile.Failed(); +} + /*}}}*/ diff --git a/apt-pkg/contrib/gpgv.h b/apt-pkg/contrib/gpgv.h new file mode 100644 index 0000000..1cabed4 --- /dev/null +++ b/apt-pkg/contrib/gpgv.h @@ -0,0 +1,89 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + Helpers to deal with gpgv better and more easily + + ##################################################################### */ + /*}}}*/ +#ifndef CONTRIB_GPGV_H +#define CONTRIB_GPGV_H + +#include <apt-pkg/macros.h> + +#include <string> +#include <vector> + + +class FileFd; + +/** \brief generates and run the command to verify a file with gpgv + * + * If File and FileSig specify the same file it is assumed that we + * deal with a clear-signed message. Note that the method will accept + * and validate files which include additional (unsigned) messages + * without complaining. Do NOT open files accepted by this method + * for reading. Use #OpenMaybeClearSignedFile to access the message + * instead to ensure you are only reading signed data. + * + * The method does not return, but has some notable exit-codes: + * 111 signals an internal error like the inability to execute gpgv, + * 112 indicates a clear-signed file which doesn't include a message, + * which can happen if APT is run while on a network requiring + * authentication before usage (e.g. in hotels) + * All other exit-codes are passed-through from gpgv. + * + * @param File is the message (unsigned or clear-signed) + * @param FileSig is the signature (detached or clear-signed) + * @param statusfd is the fd given to gpgv as --status-fd + * @param fd is used as a pipe for the standard output of gpgv + * @param key is the specific one to be used instead of using all + */ +APT_PUBLIC void ExecGPGV(std::string const &File, std::string const &FileSig, + int const &statusfd, int fd[2], std::string const &Key = "") APT_NORETURN; +inline APT_NORETURN void ExecGPGV(std::string const &File, std::string const &FileSig, + int const &statusfd = -1) { + int fd[2]; + ExecGPGV(File, FileSig, statusfd, fd); +} + +/** \brief Split an inline signature into message and signature + * + * Takes a clear-signed message and puts the first signed message + * in the content file and all signatures following it into the + * second. Unsigned messages, additional messages as well as + * whitespaces are discarded. The resulting files are suitable to + * be checked with gpgv. + * + * If a FileFd pointers is NULL it will not be used and the content + * which would have been written to it is silently discarded. + * + * The content of the split files is undefined if the splitting was + * unsuccessful. + * + * Note that trying to split an unsigned file will fail, but + * not generate an error message. + * + * @param InFile is the clear-signed file + * @param ContentFile is the FileFd the message will be written to + * @param ContentHeader is a list of all required Amored Headers for the message + * @param SignatureFile is the FileFd all signatures will be written to + * @return true if the splitting was successful, false otherwise + */ +APT_PUBLIC bool SplitClearSignedFile(std::string const &InFile, FileFd * const ContentFile, + std::vector<std::string> * const ContentHeader, FileFd * const SignatureFile); + +/** \brief open a file which might be clear-signed + * + * This method tries to extract the (signed) message of a file. + * If the file isn't signed it will just open the given filename. + * Otherwise the message is extracted to a temporary file which + * will be opened instead. + * + * @param ClearSignedFileName is the name of the file to open + * @param[out] MessageFile is the FileFd in which the file will be opened + * @return true if opening was successful, otherwise false + */ +APT_PUBLIC bool OpenMaybeClearSignedFile(std::string const &ClearSignedFileName, FileFd &MessageFile); + +#endif diff --git a/apt-pkg/contrib/hashes.cc b/apt-pkg/contrib/hashes.cc new file mode 100644 index 0000000..313b1d3 --- /dev/null +++ b/apt-pkg/contrib/hashes.cc @@ -0,0 +1,464 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + Hashes - Simple wrapper around the hash functions + + This is just used to make building the methods simpler, this is the + only interface required.. + + ##################################################################### */ + /*}}}*/ +// Include Files /*{{{*/ +#include <config.h> + +#include <apt-pkg/configuration.h> +#include <apt-pkg/fileutl.h> +#include <apt-pkg/hashes.h> +#include <apt-pkg/macros.h> +#include <apt-pkg/strutl.h> +#include <apt-pkg/tagfile-keys.h> +#include <apt-pkg/tagfile.h> + +#include <algorithm> +#include <iostream> +#include <string> +#include <assert.h> +#include <stddef.h> +#include <stdlib.h> +#include <unistd.h> + +#include <gcrypt.h> + /*}}}*/ + +static const constexpr struct HashAlgo +{ + const char *name; + int gcryAlgo; + Hashes::SupportedHashes ourAlgo; +} Algorithms[] = { + {"MD5Sum", GCRY_MD_MD5, Hashes::MD5SUM}, + {"SHA1", GCRY_MD_SHA1, Hashes::SHA1SUM}, + {"SHA256", GCRY_MD_SHA256, Hashes::SHA256SUM}, + {"SHA512", GCRY_MD_SHA512, Hashes::SHA512SUM}, +}; + +const char * HashString::_SupportedHashes[] = +{ + "SHA512", "SHA256", "SHA1", "MD5Sum", "Checksum-FileSize", NULL +}; +std::vector<HashString::HashSupportInfo> HashString::SupportedHashesInfo() +{ + return {{ + { "SHA512", pkgTagSection::Key::SHA512,"Checksums-Sha512", pkgTagSection::Key::Checksums_Sha512}, + { "SHA256", pkgTagSection::Key::SHA256, "Checksums-Sha256", pkgTagSection::Key::Checksums_Sha256}, + { "SHA1", pkgTagSection::Key::SHA1, "Checksums-Sha1", pkgTagSection::Key::Checksums_Sha1 }, + { "MD5Sum", pkgTagSection::Key::MD5sum, "Files", pkgTagSection::Key::Files }, + }}; +} + +HashString::HashString() +{ +} + +HashString::HashString(std::string Type, std::string Hash) : Type(Type), Hash(Hash) +{ +} + +HashString::HashString(std::string StringedHash) /*{{{*/ +{ + if (StringedHash.find(":") == std::string::npos) + { + // legacy: md5sum without "MD5Sum:" prefix + if (StringedHash.size() == 32) + { + Type = "MD5Sum"; + Hash = StringedHash; + } + if(_config->FindB("Debug::Hashes",false) == true) + std::clog << "HashString(string): invalid StringedHash " << StringedHash << std::endl; + return; + } + std::string::size_type pos = StringedHash.find(":"); + Type = StringedHash.substr(0,pos); + Hash = StringedHash.substr(pos+1, StringedHash.size() - pos); + + if(_config->FindB("Debug::Hashes",false) == true) + std::clog << "HashString(string): " << Type << " : " << Hash << std::endl; +} + /*}}}*/ +bool HashString::VerifyFile(std::string filename) const /*{{{*/ +{ + std::string fileHash = GetHashForFile(filename); + + if(_config->FindB("Debug::Hashes",false) == true) + std::clog << "HashString::VerifyFile: got: " << fileHash << " expected: " << toStr() << std::endl; + + return (fileHash == Hash); +} + /*}}}*/ +bool HashString::FromFile(std::string filename) /*{{{*/ +{ + // pick the strongest hash + if (Type == "") + Type = _SupportedHashes[0]; + + Hash = GetHashForFile(filename); + return true; +} + /*}}}*/ +std::string HashString::GetHashForFile(std::string filename) const /*{{{*/ +{ + std::string fileHash; + + FileFd Fd(filename, FileFd::ReadOnly); + if(strcasecmp(Type.c_str(), "MD5Sum") == 0) + { + Hashes MD5(Hashes::MD5SUM); + MD5.AddFD(Fd); + fileHash = MD5.GetHashString(Hashes::MD5SUM).Hash; + } + else if (strcasecmp(Type.c_str(), "SHA1") == 0) + { + Hashes SHA1(Hashes::SHA1SUM); + SHA1.AddFD(Fd); + fileHash = SHA1.GetHashString(Hashes::SHA1SUM).Hash; + } + else if (strcasecmp(Type.c_str(), "SHA256") == 0) + { + Hashes SHA256(Hashes::SHA256SUM); + SHA256.AddFD(Fd); + fileHash = SHA256.GetHashString(Hashes::SHA256SUM).Hash; + } + else if (strcasecmp(Type.c_str(), "SHA512") == 0) + { + Hashes SHA512(Hashes::SHA512SUM); + SHA512.AddFD(Fd); + fileHash = SHA512.GetHashString(Hashes::SHA512SUM).Hash; + } + else if (strcasecmp(Type.c_str(), "Checksum-FileSize") == 0) + strprintf(fileHash, "%llu", Fd.FileSize()); + Fd.Close(); + + return fileHash; +} + /*}}}*/ +const char** HashString::SupportedHashes() /*{{{*/ +{ + return _SupportedHashes; +} + /*}}}*/ +APT_PURE bool HashString::empty() const /*{{{*/ +{ + return (Type.empty() || Hash.empty()); +} + /*}}}*/ + +APT_PURE static bool IsConfigured(const char *name, const char *what) +{ + std::string option; + strprintf(option, "APT::Hashes::%s::%s", name, what); + return _config->FindB(option, false); +} + +APT_PURE bool HashString::usable() const /*{{{*/ +{ + return ( + (Type != "Checksum-FileSize") && + (Type != "MD5Sum") && + (Type != "SHA1") && + !IsConfigured(Type.c_str(), "Untrusted") + ); +} + /*}}}*/ +std::string HashString::toStr() const /*{{{*/ +{ + return Type + ":" + Hash; +} + /*}}}*/ +APT_PURE bool HashString::operator==(HashString const &other) const /*{{{*/ +{ + return (strcasecmp(Type.c_str(), other.Type.c_str()) == 0 && Hash == other.Hash); +} +APT_PURE bool HashString::operator!=(HashString const &other) const +{ + return !(*this == other); +} + /*}}}*/ + +bool HashStringList::usable() const /*{{{*/ +{ + if (empty() == true) + return false; + std::string const forcedType = _config->Find("Acquire::ForceHash", ""); + if (forcedType.empty() == true) + { + // See if there is at least one usable hash + return std::any_of(list.begin(), list.end(), [](auto const &hs) { return hs.usable(); }); + } + return find(forcedType) != NULL; +} + /*}}}*/ +HashString const * HashStringList::find(char const * const type) const /*{{{*/ +{ + if (type == NULL || type[0] == '\0') + { + std::string const forcedType = _config->Find("Acquire::ForceHash", ""); + if (forcedType.empty() == false) + return find(forcedType.c_str()); + for (char const * const * t = HashString::SupportedHashes(); *t != NULL; ++t) + for (std::vector<HashString>::const_iterator hs = list.begin(); hs != list.end(); ++hs) + if (strcasecmp(hs->HashType().c_str(), *t) == 0) + return &*hs; + return NULL; + } + for (std::vector<HashString>::const_iterator hs = list.begin(); hs != list.end(); ++hs) + if (strcasecmp(hs->HashType().c_str(), type) == 0) + return &*hs; + return NULL; +} + /*}}}*/ +unsigned long long HashStringList::FileSize() const /*{{{*/ +{ + HashString const * const hsf = find("Checksum-FileSize"); + if (hsf == NULL) + return 0; + std::string const hv = hsf->HashValue(); + return strtoull(hv.c_str(), NULL, 10); +} + /*}}}*/ +bool HashStringList::FileSize(unsigned long long const Size) /*{{{*/ +{ + return push_back(HashString("Checksum-FileSize", std::to_string(Size))); +} + /*}}}*/ +bool HashStringList::supported(char const * const type) /*{{{*/ +{ + for (char const * const * t = HashString::SupportedHashes(); *t != NULL; ++t) + if (strcasecmp(*t, type) == 0) + return true; + return false; +} + /*}}}*/ +bool HashStringList::push_back(const HashString &hashString) /*{{{*/ +{ + if (hashString.HashType().empty() == true || + hashString.HashValue().empty() == true || + supported(hashString.HashType().c_str()) == false) + return false; + + // ensure that each type is added only once + HashString const * const hs = find(hashString.HashType().c_str()); + if (hs != NULL) + return *hs == hashString; + + list.push_back(hashString); + return true; +} + /*}}}*/ +bool HashStringList::VerifyFile(std::string filename) const /*{{{*/ +{ + if (usable() == false) + return false; + + Hashes hashes(*this); + FileFd file(filename, FileFd::ReadOnly); + HashString const * const hsf = find("Checksum-FileSize"); + if (hsf != NULL) + { + std::string fileSize; + strprintf(fileSize, "%llu", file.FileSize()); + if (hsf->HashValue() != fileSize) + return false; + } + hashes.AddFD(file); + HashStringList const hsl = hashes.GetHashStringList(); + return hsl == *this; +} + /*}}}*/ +bool HashStringList::operator==(HashStringList const &other) const /*{{{*/ +{ + std::string const forcedType = _config->Find("Acquire::ForceHash", ""); + if (forcedType.empty() == false) + { + HashString const * const hs = find(forcedType); + HashString const * const ohs = other.find(forcedType); + if (hs == NULL || ohs == NULL) + return false; + return *hs == *ohs; + } + short matches = 0; + for (const_iterator hs = begin(); hs != end(); ++hs) + { + HashString const * const ohs = other.find(hs->HashType()); + if (ohs == NULL) + continue; + if (*hs != *ohs) + return false; + ++matches; + } + if (matches == 0) + return false; + return true; +} +bool HashStringList::operator!=(HashStringList const &other) const +{ + return !(*this == other); +} + /*}}}*/ + +// PrivateHashes /*{{{*/ +class PrivateHashes { +public: + unsigned long long FileSize; + gcry_md_hd_t hd; + + void maybeInit() + { + + // Yikes, we got to initialize libgcrypt, or we get warnings. But we + // abstract away libgcrypt in Hashes from our users - they are not + // supposed to know what the hashing backend is, so we can't force + // them to init themselves as libgcrypt folks want us to. So this + // only leaves us with this option... + if (!gcry_control(GCRYCTL_INITIALIZATION_FINISHED_P)) + { + if (!gcry_check_version(nullptr)) + { + fprintf(stderr, "libgcrypt is too old (need %s, have %s)\n", + "nullptr", gcry_check_version(NULL)); + exit(2); + } + + gcry_control(GCRYCTL_INITIALIZATION_FINISHED, 0); + } + } + + explicit PrivateHashes(unsigned int const CalcHashes) : FileSize(0) + { + maybeInit(); + gcry_md_open(&hd, 0, 0); + for (auto & Algo : Algorithms) + { + if ((CalcHashes & Algo.ourAlgo) == Algo.ourAlgo) + gcry_md_enable(hd, Algo.gcryAlgo); + } + } + + explicit PrivateHashes(HashStringList const &Hashes) : FileSize(0) { + maybeInit(); + gcry_md_open(&hd, 0, 0); + for (auto & Algo : Algorithms) + { + if (not Hashes.usable() || Hashes.find(Algo.name) != NULL) + gcry_md_enable(hd, Algo.gcryAlgo); + } + } + ~PrivateHashes() + { + gcry_md_close(hd); + } +}; + /*}}}*/ +// Hashes::Add* - Add the contents of data or FD /*{{{*/ +bool Hashes::Add(const unsigned char * const Data, unsigned long long const Size) +{ + if (Size != 0) + { + gcry_md_write(d->hd, Data, Size); + d->FileSize += Size; + } + return true; +} +bool Hashes::AddFD(int const Fd,unsigned long long Size) +{ + unsigned char Buf[APT_BUFFER_SIZE]; + bool const ToEOF = (Size == UntilEOF); + while (Size != 0 || ToEOF) + { + decltype(Size) n = sizeof(Buf); + if (!ToEOF) n = std::min(Size, n); + ssize_t const Res = read(Fd,Buf,n); + if (Res < 0 || (!ToEOF && Res != (ssize_t) n)) // error, or short read + return false; + if (ToEOF && Res == 0) // EOF + break; + Size -= Res; + if (Add(Buf, Res) == false) + return false; + } + return true; +} +bool Hashes::AddFD(FileFd &Fd,unsigned long long Size) +{ + unsigned char Buf[APT_BUFFER_SIZE]; + bool const ToEOF = (Size == 0); + while (Size != 0 || ToEOF) + { + decltype(Size) n = sizeof(Buf); + if (!ToEOF) n = std::min(Size, n); + decltype(Size) a = 0; + if (Fd.Read(Buf, n, &a) == false) // error + return false; + if (ToEOF == false) + { + if (a != n) // short read + return false; + } + else if (a == 0) // EOF + break; + Size -= a; + if (Add(Buf, a) == false) + return false; + } + return true; +} + /*}}}*/ + +static APT_PURE std::string HexDigest(gcry_md_hd_t hd, int algo) +{ + char Conv[16] = + {'0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'a', 'b', + 'c', 'd', 'e', 'f'}; + + auto Size = gcry_md_get_algo_dlen(algo); + assert(Size <= 512/8); + char Result[((Size)*2) + 1]; + Result[(Size)*2] = 0; + + auto Sum = gcry_md_read(hd, algo); + + // Convert each char into two letters + size_t J = 0; + size_t I = 0; + for (; I != (Size)*2; J++, I += 2) + { + Result[I] = Conv[Sum[J] >> 4]; + Result[I + 1] = Conv[Sum[J] & 0xF]; + } + return std::string(Result); +}; + +HashStringList Hashes::GetHashStringList() +{ + HashStringList hashes; + for (auto & Algo : Algorithms) + if (gcry_md_is_enabled(d->hd, Algo.gcryAlgo)) + hashes.push_back(HashString(Algo.name, HexDigest(d->hd, Algo.gcryAlgo))); + hashes.FileSize(d->FileSize); + + return hashes; +} + +HashString Hashes::GetHashString(SupportedHashes hash) +{ + for (auto & Algo : Algorithms) + if (hash == Algo.ourAlgo) + return HashString(Algo.name, HexDigest(d->hd, Algo.gcryAlgo)); + + abort(); +} +Hashes::Hashes() : d(new PrivateHashes(~0)) { } +Hashes::Hashes(unsigned int const Hashes) : d(new PrivateHashes(Hashes)) {} +Hashes::Hashes(HashStringList const &Hashes) : d(new PrivateHashes(Hashes)) {} +Hashes::~Hashes() { delete d; } diff --git a/apt-pkg/contrib/hashes.h b/apt-pkg/contrib/hashes.h new file mode 100644 index 0000000..e259b4e --- /dev/null +++ b/apt-pkg/contrib/hashes.h @@ -0,0 +1,222 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + Hashes - Simple wrapper around the hash functions + + This is just used to make building the methods simpler, this is the + only interface required.. + + ##################################################################### */ + /*}}}*/ +#ifndef APTPKG_HASHES_H +#define APTPKG_HASHES_H + +#include <apt-pkg/macros.h> + +#ifdef APT_COMPILING_APT +#include <apt-pkg/string_view.h> +#include <apt-pkg/tagfile-keys.h> +#endif + +#include <cstring> +#include <string> +#include <vector> + + + +class FileFd; + +// helper class that contains hash function name +// and hash +class APT_PUBLIC HashString +{ + protected: + std::string Type; + std::string Hash; + static const char * _SupportedHashes[10]; + + // internal helper + std::string GetHashForFile(std::string filename) const; + + public: + HashString(std::string Type, std::string Hash); + explicit HashString(std::string StringedHashString); // init from str as "type:hash" + HashString(); + + // get hash type used + std::string HashType() const { return Type; }; + std::string HashValue() const { return Hash; }; + + // verify the given filename against the currently loaded hash + bool VerifyFile(std::string filename) const; + + // generate a hash string from the given filename + bool FromFile(std::string filename); + + + // helper + std::string toStr() const; // convert to str as "type:hash" + bool empty() const; + bool usable() const; + bool operator==(HashString const &other) const; + bool operator!=(HashString const &other) const; + + // return the list of hashes we support + static APT_PURE const char** SupportedHashes(); +#ifdef APT_COMPILING_APT + struct APT_HIDDEN HashSupportInfo { + APT::StringView name; + pkgTagSection::Key namekey; + APT::StringView chksumsname; + pkgTagSection::Key chksumskey; + }; + APT_HIDDEN static std::vector<HashSupportInfo> SupportedHashesInfo(); +#endif +}; + +class APT_PUBLIC HashStringList +{ + public: + /** find best hash if no specific one is requested + * + * @param type of the checksum to return, can be \b NULL + * @return If type is \b NULL (or the empty string) it will + * return the 'best' hash; otherwise the hash which was + * specifically requested. If no hash is found \b NULL will be returned. + */ + HashString const * find(char const * const type) const; + HashString const * find(std::string const &type) const { return find(type.c_str()); } + + /** finds the filesize hash and returns it as number + * + * @return beware: if the size isn't known we return \b 0 here, + * just like we would do for an empty file. If that is a problem + * for you have to get the size manually out of the list. + */ + unsigned long long FileSize() const; + + /** sets the filesize hash + * + * @param Size of the file + * @return @see #push_back + */ + bool FileSize(unsigned long long const Size); + + /** check if the given hash type is supported + * + * @param type to check + * @return true if supported, otherwise false + */ + static APT_PURE bool supported(char const * const type); + /** add the given #HashString to the list + * + * @param hashString to add + * @return true if the hash is added because it is supported and + * not already a different hash of the same type included, otherwise false + */ + bool push_back(const HashString &hashString); + /** @return size of the list of HashStrings */ + size_t size() const { return list.size(); } + + /** verify file against all hashes in the list + * + * @param filename to verify + * @return true if the file matches the hashsum, otherwise false + */ + bool VerifyFile(std::string filename) const; + + /** is the list empty ? + * + * @return \b true if the list is empty, otherwise \b false + */ + bool empty() const { return list.empty(); } + + /** has the list at least one good entry + * + * similar to #empty, but handles forced hashes. + * + * @return if no hash is forced, same result as #empty, + * if one is forced \b true if this has is available, \b false otherwise + */ + bool usable() const; + + typedef std::vector<HashString>::const_iterator const_iterator; + + /** iterator to the first element */ + const_iterator begin() const { return list.begin(); } + + /** iterator to the end element */ + const_iterator end() const { return list.end(); } + + /** start fresh with a clear list */ + void clear() { list.clear(); } + + /** compare two HashStringList for similarity. + * + * Two lists are similar if at least one hashtype is in both lists + * and the hashsum matches. All hashes are checked by default, + * if one doesn't match false is returned regardless of how many + * matched before. If a hash is forced, only this hash is compared, + * all others are ignored. + */ + bool operator==(HashStringList const &other) const; + bool operator!=(HashStringList const &other) const; + + HashStringList() {} + + // simplifying API-compatibility constructors + explicit HashStringList(std::string const &hash) { + if (hash.empty() == false) + list.push_back(HashString(hash)); + } + explicit HashStringList(char const * const hash) { + if (hash != NULL && hash[0] != '\0') + list.push_back(HashString(hash)); + } + + private: + std::vector<HashString> list; +}; + +class PrivateHashes; +class APT_PUBLIC Hashes +{ + PrivateHashes * const d; + public: + static const int UntilEOF = 0; + + bool Add(const unsigned char * const Data, unsigned long long const Size) APT_NONNULL(2); + inline bool Add(const char * const Data) APT_NONNULL(2) + {return Add(reinterpret_cast<unsigned char const *>(Data),strlen(Data));}; + inline bool Add(const char *const Data, unsigned long long const Size) APT_NONNULL(2) + { + return Add(reinterpret_cast<unsigned char const *>(Data), Size); + }; + inline bool Add(const unsigned char * const Beg,const unsigned char * const End) APT_NONNULL(2,3) + {return Add(Beg,End-Beg);}; + + enum SupportedHashes { MD5SUM = (1 << 0), SHA1SUM = (1 << 1), SHA256SUM = (1 << 2), + SHA512SUM = (1 << 3) }; + bool AddFD(int const Fd,unsigned long long Size = 0); + bool AddFD(FileFd &Fd,unsigned long long Size = 0); + + HashStringList GetHashStringList(); + + /** Get a specific hash. It is an error to use a hash that was not hashes */ + HashString GetHashString(SupportedHashes hash); + + /** create a Hashes object to calculate all supported hashes + * + * If ALL is too much, you can limit which Hashes are calculated + * with the following other constructors which mention explicitly + * which hashes to generate. */ + Hashes(); + /** @param Hashes bitflag composed of #SupportedHashes */ + explicit Hashes(unsigned int const Hashes); + /** @param Hashes is a list of hashes */ + explicit Hashes(HashStringList const &Hashes); + virtual ~Hashes(); +}; + +#endif diff --git a/apt-pkg/contrib/header-is-private.h b/apt-pkg/contrib/header-is-private.h new file mode 100644 index 0000000..201a40c --- /dev/null +++ b/apt-pkg/contrib/header-is-private.h @@ -0,0 +1,3 @@ +#ifndef APT_COMPILING_APT +#error Internal header without ABI stability. Should only be used by apt itself! +#endif diff --git a/apt-pkg/contrib/macros.h b/apt-pkg/contrib/macros.h new file mode 100644 index 0000000..c08cd24 --- /dev/null +++ b/apt-pkg/contrib/macros.h @@ -0,0 +1,131 @@ +// -*- mode: cpp; mode: fold -*- +// SPDX-License-Identifier: GPL-2.0+ +// Description /*{{{*/ +/* ###################################################################### + + Macros Header - Various useful macro definitions + + This file had this historic note, but now includes further changes + under the GPL-2.0+: + + This source is placed in the Public Domain, do with it what you will + It was originally written by Brian C. White. + + ##################################################################### */ + /*}}}*/ +// Private header +#ifndef MACROS_H +#define MACROS_H + +/* Useful count macro, use on an array of things and it will return the + number of items in the array */ +#define APT_ARRAY_SIZE(a) (sizeof(a) / sizeof(a[0])) + +// Flag Macros +#define FLAG(f) (1L << (f)) +#define SETFLAG(v,f) ((v) |= FLAG(f)) +#define CLRFLAG(v,f) ((v) &=~FLAG(f)) +#define CHKFLAG(v,f) ((v) & FLAG(f) ? true : false) + +#ifdef __GNUC__ +#define APT_GCC_VERSION (__GNUC__ << 8 | __GNUC_MINOR__) +#else +#define APT_GCC_VERSION 0 +#endif + +#ifdef APT_COMPILING_APT +/* likely() and unlikely() can be used to mark boolean expressions + as (not) likely true which will help the compiler to optimise */ +#if APT_GCC_VERSION >= 0x0300 + #define likely(x) __builtin_expect (!!(x), 1) + #define unlikely(x) __builtin_expect (!!(x), 0) +#else + #define likely(x) (x) + #define unlikely(x) (x) +#endif +#endif + +#if APT_GCC_VERSION >= 0x0300 + #define APT_DEPRECATED __attribute__ ((deprecated)) + #define APT_DEPRECATED_MSG(X) __attribute__ ((deprecated(X))) + // __attribute__((const)) is too dangerous for us, we end up using it wrongly + #define APT_PURE __attribute__((pure)) + #define APT_NORETURN __attribute__((noreturn)) + #define APT_PRINTF(n) __attribute__((format(printf, n, n + 1))) + #define APT_WEAK __attribute__((weak)); + #define APT_UNUSED __attribute__((unused)) +#else + #define APT_DEPRECATED + #define APT_DEPRECATED_MSG + #define APT_PURE + #define APT_NORETURN + #define APT_PRINTF(n) + #define APT_WEAK + #define APT_UNUSED +#endif + +#if APT_GCC_VERSION > 0x0302 + #define APT_NONNULL(...) __attribute__((nonnull(__VA_ARGS__))) + #define APT_MUSTCHECK __attribute__((warn_unused_result)) +#else + #define APT_NONNULL(...) + #define APT_MUSTCHECK +#endif + +#if APT_GCC_VERSION >= 0x0400 + #define APT_SENTINEL __attribute__((sentinel)) + #define APT_PUBLIC __attribute__ ((visibility ("default"))) + #define APT_HIDDEN __attribute__ ((visibility ("hidden"))) +#else + #define APT_SENTINEL + #define APT_PUBLIC + #define APT_HIDDEN +#endif + +// cold functions are unlikely() to be called +#if APT_GCC_VERSION >= 0x0403 + #define APT_COLD __attribute__ ((__cold__)) + #define APT_HOT __attribute__ ((__hot__)) +#else + #define APT_COLD + #define APT_HOT +#endif + + +#if __GNUC__ >= 4 + #define APT_IGNORE_DEPRECATED_PUSH \ + _Pragma("GCC diagnostic push") \ + _Pragma("GCC diagnostic ignored \"-Wdeprecated-declarations\"") + #define APT_IGNORE_DEPRECATED_POP \ + _Pragma("GCC diagnostic pop") + /* gcc has various problems with this shortcut, so prefer the long form */ + #define APT_IGNORE_DEPRECATED(XXX) \ + APT_IGNORE_DEPRECATED_PUSH \ + XXX \ + APT_IGNORE_DEPRECATED_POP +#else + #define APT_IGNORE_DEPRECATED_PUSH + #define APT_IGNORE_DEPRECATED_POP + #define APT_IGNORE_DEPRECATED(XXX) XXX +#endif + +#if __cplusplus >= 201103L + #define APT_OVERRIDE override +#else + #define APT_OVERRIDE /* no c++11 standard */ +#endif + +// These lines are extracted by the makefiles and the buildsystem +// Increasing MAJOR or MINOR results in the need of recompiling all +// reverse-dependencies of libapt-pkg against the new SONAME. +// Non-ABI-Breaks should only increase RELEASE number. +// See also buildlib/libversion.mak +#define APT_PKG_MAJOR 6 +#define APT_PKG_MINOR 0 +#define APT_PKG_RELEASE 0 +#define APT_PKG_ABI ((APT_PKG_MAJOR * 100) + APT_PKG_MINOR) + +/* Should be a multiple of the common page size (4096) */ +static constexpr unsigned long long APT_BUFFER_SIZE = 64 * 1024; + +#endif diff --git a/apt-pkg/contrib/mmap.cc b/apt-pkg/contrib/mmap.cc new file mode 100644 index 0000000..0568e1c --- /dev/null +++ b/apt-pkg/contrib/mmap.cc @@ -0,0 +1,504 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + MMap Class - Provides 'real' mmap or a faked mmap using read(). + + MMap cover class. + + Some broken versions of glibc2 (libc6) have a broken definition + of mmap that accepts a char * -- all other systems (and libc5) use + void *. We can't safely do anything here that would be portable, so + libc6 generates warnings -- which should be errors, g++ isn't properly + strict. + + ##################################################################### */ + /*}}}*/ +// Include Files /*{{{*/ +#define _DEFAULT_SOURCE +#include <config.h> + +#include <apt-pkg/error.h> +#include <apt-pkg/fileutl.h> +#include <apt-pkg/macros.h> +#include <apt-pkg/mmap.h> + +#include <cstdint> +#include <cstring> +#include <string> +#include <errno.h> +#include <stdlib.h> +#include <unistd.h> + +#include <apti18n.h> + /*}}}*/ + +// MMap::MMap - Constructor /*{{{*/ +// --------------------------------------------------------------------- +/* */ +MMap::MMap(FileFd &F,unsigned long Flags) : Flags(Flags), iSize(0), + Base(nullptr), SyncToFd(nullptr) +{ + if ((Flags & NoImmMap) != NoImmMap) + Map(F); +} + /*}}}*/ +// MMap::MMap - Constructor /*{{{*/ +// --------------------------------------------------------------------- +/* */ +MMap::MMap(unsigned long Flags) : Flags(Flags), iSize(0), + Base(nullptr), SyncToFd(nullptr) +{ +} + /*}}}*/ +// MMap::~MMap - Destructor /*{{{*/ +// --------------------------------------------------------------------- +/* */ +MMap::~MMap() +{ + Close(); +} + /*}}}*/ +// MMap::Map - Perform the mapping /*{{{*/ +// --------------------------------------------------------------------- +/* */ +bool MMap::Map(FileFd &Fd) +{ + iSize = Fd.Size(); + + // Set the permissions. + int Prot = PROT_READ; + int Map = MAP_SHARED; + if ((Flags & ReadOnly) != ReadOnly) + Prot |= PROT_WRITE; + if ((Flags & Public) != Public) + Map = MAP_PRIVATE; + + if (iSize == 0) + return _error->Error(_("Can't mmap an empty file")); + + // We can't mmap compressed fd's directly, so we need to read it completely + if (Fd.IsCompressed() == true) + { + if ((Flags & ReadOnly) != ReadOnly) + return _error->Error("Compressed file %s can only be mapped readonly", Fd.Name().c_str()); + Base = malloc(iSize); + if (unlikely(Base == nullptr)) + return _error->Errno("MMap-compressed-malloc", _("Couldn't make mmap of %llu bytes"), iSize); + SyncToFd = new FileFd(); + if (Fd.Seek(0L) == false || Fd.Read(Base, iSize) == false) + return _error->Error("Compressed file %s can't be read into mmap", Fd.Name().c_str()); + return true; + } + + // Map it. + Base = (Flags & Fallback) ? MAP_FAILED : mmap(0,iSize,Prot,Map,Fd.Fd(),0); + if (Base == MAP_FAILED) + { + if (errno == ENODEV || errno == EINVAL || (Flags & Fallback)) + { + // The filesystem doesn't support this particular kind of mmap. + // So we allocate a buffer and read the whole file into it. + if ((Flags & ReadOnly) == ReadOnly) + { + // for readonly, we don't need sync, so make it simple + Base = malloc(iSize); + if (unlikely(Base == nullptr)) + return _error->Errno("MMap-malloc", _("Couldn't make mmap of %llu bytes"), iSize); + SyncToFd = new FileFd(); + return Fd.Read(Base, iSize); + } + // FIXME: Writing to compressed fd's ? + int const dupped_fd = dup(Fd.Fd()); + if (dupped_fd == -1) + return _error->Errno("mmap", _("Couldn't duplicate file descriptor %i"), Fd.Fd()); + + Base = calloc(iSize, 1); + if (unlikely(Base == nullptr)) + return _error->Errno("MMap-calloc", _("Couldn't make mmap of %llu bytes"), iSize); + SyncToFd = new FileFd (dupped_fd); + if (!SyncToFd->Seek(0L) || !SyncToFd->Read(Base, iSize)) + return false; + } + else + return _error->Errno("MMap-mmap", _("Couldn't make mmap of %llu bytes"), iSize); + } + + return true; +} + /*}}}*/ +// MMap::Close - Close the map /*{{{*/ +// --------------------------------------------------------------------- +/* */ +bool MMap::Close(bool DoSync) +{ + if ((Flags & UnMapped) == UnMapped || validData() == false || iSize == 0) + return true; + + if (DoSync == true) + Sync(); + + if (SyncToFd != NULL) + { + free(Base); + delete SyncToFd; + SyncToFd = NULL; + } + else + { + if (munmap((char *)Base, iSize) != 0) + _error->WarningE("mmap", _("Unable to close mmap")); + } + + iSize = 0; + Base = 0; + return true; +} + /*}}}*/ +// MMap::Sync - Synchronize the map with the disk /*{{{*/ +// --------------------------------------------------------------------- +/* This is done in synchronous mode - the docs indicate that this will + not return till all IO is complete */ +bool MMap::Sync() +{ + if ((Flags & UnMapped) == UnMapped) + return true; + + if ((Flags & ReadOnly) != ReadOnly) + { + if (SyncToFd != NULL) + { + if (!SyncToFd->Seek(0) || !SyncToFd->Write(Base, iSize)) + return false; + } + else + { +#ifdef _POSIX_SYNCHRONIZED_IO + if (msync((char *)Base, iSize, MS_SYNC) < 0) + return _error->Errno("msync", _("Unable to synchronize mmap")); +#endif + } + } + return true; +} + /*}}}*/ +// MMap::Sync - Synchronize a section of the file to disk /*{{{*/ +// --------------------------------------------------------------------- +/* */ +bool MMap::Sync(unsigned long Start,unsigned long Stop) +{ + if ((Flags & UnMapped) == UnMapped) + return true; + + if ((Flags & ReadOnly) != ReadOnly) + { + if (SyncToFd != 0) + { + if (!SyncToFd->Seek(0) || + !SyncToFd->Write (((char *)Base)+Start, Stop-Start)) + return false; + } + else + { +#ifdef _POSIX_SYNCHRONIZED_IO + unsigned long long const PSize = sysconf(_SC_PAGESIZE); + if (msync((char *)Base+(Start/PSize)*PSize, Stop - Start, MS_SYNC) < 0) + return _error->Errno("msync", _("Unable to synchronize mmap")); +#endif + } + } + return true; +} + /*}}}*/ + +// DynamicMMap::DynamicMMap - Constructor /*{{{*/ +// --------------------------------------------------------------------- +/* */ +DynamicMMap::DynamicMMap(FileFd &F,unsigned long Flags,unsigned long const &Workspace, + unsigned long const &Grow, unsigned long const &Limit) : + MMap(F,Flags | NoImmMap), Fd(&F), WorkSpace(Workspace), + GrowFactor(Grow), Limit(Limit) +{ + // disable Moveable if we don't grow + if (Grow == 0) + this->Flags &= ~Moveable; + +#ifndef __linux__ + // kfreebsd doesn't have mremap, so we use the fallback + if ((this->Flags & Moveable) == Moveable) + this->Flags |= Fallback; +#endif + + unsigned long long EndOfFile = Fd->Size(); + if (EndOfFile > WorkSpace) + WorkSpace = EndOfFile; + else if(WorkSpace > 0) + { + Fd->Seek(WorkSpace - 1); + char C = 0; + Fd->Write(&C,sizeof(C)); + } + + Map(F); + iSize = EndOfFile; +} + /*}}}*/ +// DynamicMMap::DynamicMMap - Constructor for a non-file backed map /*{{{*/ +// --------------------------------------------------------------------- +/* We try here to use mmap to reserve some space - this is much more + cooler than the fallback solution to simply allocate a char array + and could come in handy later than we are able to grow such an mmap */ +DynamicMMap::DynamicMMap(unsigned long Flags,unsigned long const &WorkSpace, + unsigned long const &Grow, unsigned long const &Limit) : + MMap(Flags | NoImmMap | UnMapped), Fd(0), WorkSpace(WorkSpace), + GrowFactor(Grow), Limit(Limit) +{ + // disable Moveable if we don't grow + if (Grow == 0) + this->Flags &= ~Moveable; + +#ifndef __linux__ + // kfreebsd doesn't have mremap, so we use the fallback + if ((this->Flags & Moveable) == Moveable) + this->Flags |= Fallback; +#endif + +#ifdef _POSIX_MAPPED_FILES + if ((this->Flags & Fallback) != Fallback) { + // Set the permissions. + int Prot = PROT_READ; +#ifdef MAP_ANONYMOUS + int Map = MAP_PRIVATE | MAP_ANONYMOUS; +#else + int Map = MAP_PRIVATE | MAP_ANON; +#endif + if ((this->Flags & ReadOnly) != ReadOnly) + Prot |= PROT_WRITE; + if ((this->Flags & Public) == Public) +#ifdef MAP_ANONYMOUS + Map = MAP_SHARED | MAP_ANONYMOUS; +#else + Map = MAP_SHARED | MAP_ANON; +#endif + + // use anonymous mmap() to get the memory + Base = (unsigned char*) mmap(0, WorkSpace, Prot, Map, -1, 0); + + if(Base == MAP_FAILED) + _error->Errno("DynamicMMap",_("Couldn't make mmap of %lu bytes"),WorkSpace); + + iSize = 0; + return; + } +#endif + // fallback to a static allocated space + Base = calloc(WorkSpace, 1); + iSize = 0; +} + /*}}}*/ +// DynamicMMap::~DynamicMMap - Destructor /*{{{*/ +// --------------------------------------------------------------------- +/* We truncate the file to the size of the memory data set */ +DynamicMMap::~DynamicMMap() +{ + if (Fd == 0) + { + if (validData() == false) + return; +#ifdef _POSIX_MAPPED_FILES + munmap(Base, WorkSpace); +#else + free(Base); +#endif + return; + } + + unsigned long long EndOfFile = iSize; + iSize = WorkSpace; + Close(false); + if(ftruncate(Fd->Fd(),EndOfFile) < 0) + _error->Errno("ftruncate", _("Failed to truncate file")); +} + /*}}}*/ +// DynamicMMap::RawAllocate - Allocate a raw chunk of unaligned space /*{{{*/ +// --------------------------------------------------------------------- +/* This allocates a block of memory aligned to the given size */ +unsigned long DynamicMMap::RawAllocate(unsigned long long Size,unsigned long Aln) +{ + unsigned long long Result = iSize; + if (Aln != 0) + Result += Aln - (iSize%Aln); + + iSize = Result + Size; + + // try to grow the buffer + while(Result + Size > WorkSpace) + { + if(!Grow()) + { + _error->Fatal(_("Dynamic MMap ran out of room. Please increase the size " + "of APT::Cache-Start. Current value: %lu. (man 5 apt.conf)"), WorkSpace); + return 0; + } + } + return Result; +} + /*}}}*/ +// DynamicMMap::Allocate - Pooled aligned allocation /*{{{*/ +// --------------------------------------------------------------------- +/* This allocates an Item of size ItemSize so that it is aligned to its + size in the file. */ +unsigned long DynamicMMap::Allocate(unsigned long ItemSize) +{ + if (unlikely(ItemSize == 0)) + { + _error->Fatal("Can't allocate an item of size zero"); + return 0; + } + + // Look for a matching pool entry + Pool *I; + for (I = Pools; I != Pools + PoolCount; ++I) + { + if (I->ItemSize == ItemSize) + break; + } + // No pool is allocated, use an unallocated one. + if (unlikely(I == Pools + PoolCount)) + { + for (I = Pools; I != Pools + PoolCount; ++I) + { + if (I->ItemSize == 0) + break; + } + // Woops, we ran out, the calling code should allocate more. + if (I == Pools + PoolCount) + { + _error->Error("Ran out of allocation pools"); + return 0; + } + + I->ItemSize = ItemSize; + I->Count = 0; + } + + unsigned long Result = 0; + // Out of space, allocate some more + if (I->Count == 0) + { + const unsigned long size = 20*1024; + I->Count = size/ItemSize; + Pool* oldPools = Pools; + _error->PushToStack(); + Result = RawAllocate(size,ItemSize); + bool const newError = _error->PendingError(); + _error->MergeWithStack(); + if (Pools != oldPools) + I = Pools + (I - oldPools); + + // Does the allocation failed ? + if (Result == 0 && newError) + return 0; + I->Start = Result; + } + else + Result = I->Start; + + I->Count--; + I->Start += ItemSize; + return Result/ItemSize; +} + /*}}}*/ +// DynamicMMap::WriteString - Write a string to the file /*{{{*/ +// --------------------------------------------------------------------- +/* Strings are aligned to 16 bytes */ +unsigned long DynamicMMap::WriteString(const char *String, + unsigned long Len) +{ + if (Len == std::numeric_limits<unsigned long>::max()) + Len = strlen(String); + + _error->PushToStack(); + unsigned long Result = RawAllocate(Len+1+sizeof(uint16_t),sizeof(uint16_t)); + bool const newError = _error->PendingError(); + _error->MergeWithStack(); + + if (Base == NULL || (Result == 0 && newError)) + return 0; + + if (Len >= std::numeric_limits<uint16_t>::max()) + abort(); + + uint16_t LenToWrite = Len; + memcpy((char *)Base + Result, &LenToWrite, sizeof(LenToWrite)); + Result += + sizeof(LenToWrite); + + memcpy((char *)Base + Result,String,Len); + ((char *)Base)[Result + Len] = 0; + return Result; +} + /*}}}*/ +// DynamicMMap::Grow - Grow the mmap /*{{{*/ +// --------------------------------------------------------------------- +/* This method is a wrapper around different methods to (try to) grow + a mmap (or our char[]-fallback). Encounterable environments: + 1. Moveable + !Fallback + linux -> mremap with MREMAP_MAYMOVE + 2. Moveable + !Fallback + !linux -> not possible (forbidden by constructor) + 3. Moveable + Fallback -> realloc + 4. !Moveable + !Fallback + linux -> mremap alone - which will fail in 99,9% + 5. !Moveable + !Fallback + !linux -> not possible (forbidden by constructor) + 6. !Moveable + Fallback -> not possible + [ While Moveable and Fallback stands for the equally named flags and + "linux" indicates a linux kernel instead of a freebsd kernel. ] + So what you can see here is, that a MMAP which want to be growable need + to be moveable to have a real chance but that this method will at least try + the nearly impossible 4 to grow it before it finally give up: Never say never. */ +bool DynamicMMap::Grow() { + if (Limit != 0 && WorkSpace >= Limit) + return _error->Error(_("Unable to increase the size of the MMap as the " + "limit of %lu bytes is already reached."), Limit); + if (GrowFactor <= 0) + return _error->Error(_("Unable to increase size of the MMap as automatic growing is disabled by user.")); + + unsigned long long const newSize = WorkSpace + GrowFactor; + + if(Fd != 0) { + Fd->Seek(newSize - 1); + char C = 0; + Fd->Write(&C,sizeof(C)); + } + + unsigned long const poolOffset = Pools - ((Pool*) Base); + + if ((Flags & Fallback) != Fallback) { +#if defined(_POSIX_MAPPED_FILES) && defined(__linux__) + #ifdef MREMAP_MAYMOVE + + if ((Flags & Moveable) == Moveable) + Base = mremap(Base, WorkSpace, newSize, MREMAP_MAYMOVE); + else + #endif + Base = mremap(Base, WorkSpace, newSize, 0); + + if(Base == MAP_FAILED) + return false; +#else + return false; +#endif + } else { + if ((Flags & Moveable) != Moveable) + return false; + + Base = realloc(Base, newSize); + if (Base == NULL) + return false; + else + /* Set new memory to 0 */ + memset((char*)Base + WorkSpace, 0, newSize - WorkSpace); + } + + Pools =(Pool*) Base + poolOffset; + WorkSpace = newSize; + return true; +} + /*}}}*/ diff --git a/apt-pkg/contrib/mmap.h b/apt-pkg/contrib/mmap.h new file mode 100644 index 0000000..fe834d5 --- /dev/null +++ b/apt-pkg/contrib/mmap.h @@ -0,0 +1,120 @@ +// -*- mode: cpp; mode: fold -*- +// SPDX-License-Identifier: GPL-2.0+ +// Description /*{{{*/ +/* ###################################################################### + + MMap Class - Provides 'real' mmap or a faked mmap using read(). + + The purpose of this code is to provide a generic way for clients to + access the mmap function. In environments that do not support mmap + from file fd's this function will use read and normal allocated + memory. + + Writing to a public mmap will always fully commit all changes when the + class is deleted. Ie it will rewrite the file, unless it is readonly + + The DynamicMMap class is used to help the on-disk data structure + generators. It provides a large allocated workspace and members + to allocate space from the workspace in an efficient fashion. + + This file had this historic note, but now includes further changes + under the GPL-2.0+: + + This source is placed in the Public Domain, do with it what you will + It was originally written by Jason Gunthorpe. + + ##################################################################### */ + /*}}}*/ +#ifndef PKGLIB_MMAP_H +#define PKGLIB_MMAP_H + +#include <string> +#include <limits> + +#include <sys/mman.h> + + +class FileFd; + +/* This should be a 32 bit type, larger types use too much ram and smaller + types are too small. Where ever possible 'unsigned long' should be used + instead of this internal type */ +typedef unsigned int map_ptrloc; + +class MMap +{ + protected: + + unsigned long Flags; + unsigned long long iSize; + void *Base; + + // In case mmap can not be used, we keep a dup of the file + // descriptor that should have been mmaped so that we can write to + // the file in Sync(). + FileFd *SyncToFd; + + bool Map(FileFd &Fd); + bool Close(bool DoSync = true); + + public: + + enum OpenFlags {NoImmMap = (1<<0),Public = (1<<1),ReadOnly = (1<<2), + UnMapped = (1<<3), Moveable = (1<<4), Fallback = (1 << 5)}; + + // Simple accessors + inline operator void *() {return Base;}; + inline void *Data() {return Base;}; + inline unsigned long long Size() {return iSize;}; + inline void AddSize(unsigned long long const size) {iSize += size;}; + inline bool validData() const { return Base != MAP_FAILED && Base != 0; }; + + // File manipulators + bool Sync(); + bool Sync(unsigned long Start,unsigned long Stop); + + MMap(FileFd &F,unsigned long Flags); + explicit MMap(unsigned long Flags); + virtual ~MMap(); +}; + +class DynamicMMap : public MMap +{ + public: + + // This is the allocation pool structure + struct Pool + { + unsigned long ItemSize; + unsigned long Start; + unsigned long Count; + }; + + protected: + + FileFd *Fd; + unsigned long WorkSpace; + unsigned long const GrowFactor; + unsigned long const Limit; + Pool *Pools; + unsigned int PoolCount; + + bool Grow(); + + public: + + // Allocation + unsigned long RawAllocate(unsigned long long Size,unsigned long Aln = 0); + unsigned long Allocate(unsigned long ItemSize); + unsigned long WriteString(const char *String,unsigned long Len = std::numeric_limits<unsigned long>::max()); + inline unsigned long WriteString(const std::string &S) {return WriteString(S.c_str(),S.length());}; + void UsePools(Pool &P,unsigned int Count) {Pools = &P; PoolCount = Count;}; + + DynamicMMap(FileFd &F,unsigned long Flags,unsigned long const &WorkSpace = 2*1024*1024, + unsigned long const &Grow = 1024*1024, unsigned long const &Limit = 0); + DynamicMMap(unsigned long Flags,unsigned long const &WorkSpace = 2*1024*1024, + unsigned long const &Grow = 1024*1024, unsigned long const &Limit = 0); + virtual ~DynamicMMap(); +}; + +#endif diff --git a/apt-pkg/contrib/netrc.cc b/apt-pkg/contrib/netrc.cc new file mode 100644 index 0000000..ec694da --- /dev/null +++ b/apt-pkg/contrib/netrc.cc @@ -0,0 +1,173 @@ +// -*- mode: cpp; mode: fold -*- +// SPDX-License-Identifier: GPL-2.0+ +// Description /*{{{*/ +/* ###################################################################### + + netrc file parser - returns the login and password of a give host in + a specified netrc-type file + + This file had this historic note, but now includes further changes + under the GPL-2.0+: + + Originally written by Daniel Stenberg, <daniel@haxx.se>, et al. and + placed into the Public Domain, do with it what you will. + + ##################################################################### */ + /*}}}*/ +#include <config.h> +#include <apti18n.h> + +#include <apt-pkg/configuration.h> +#include <apt-pkg/error.h> +#include <apt-pkg/fileutl.h> +#include <apt-pkg/strutl.h> + +#include <iostream> + +#include "netrc.h" + +/* Get user and password from .netrc when given a machine name */ +bool MaybeAddAuth(FileFd &NetRCFile, URI &Uri) +{ + if (Uri.User.empty() == false || Uri.Password.empty() == false) + return true; + if (NetRCFile.IsOpen() == false || NetRCFile.Failed()) + return false; + auto const Debug = _config->FindB("Debug::Acquire::netrc", false); + + std::string lookfor; + if (Uri.Port != 0) + strprintf(lookfor, "%s:%i%s", Uri.Host.c_str(), Uri.Port, Uri.Path.c_str()); + else + lookfor.append(Uri.Host).append(Uri.Path); + + enum + { + NO, + MACHINE, + GOOD_MACHINE, + LOGIN, + PASSWORD + } active_token = NO; + std::string line; + while (NetRCFile.Eof() == false || line.empty() == false) + { + bool protocolSpecified = false; + + if (line.empty()) + { + if (NetRCFile.ReadLine(line) == false) + break; + else if (line.empty()) + continue; + } + auto tokenend = line.find_first_of("\t "); + std::string token; + if (tokenend != std::string::npos) + { + token = line.substr(0, tokenend); + line.erase(0, tokenend + 1); + } + else + std::swap(line, token); + if (token.empty()) + continue; + switch (active_token) + { + case NO: + if (token == "machine") + active_token = MACHINE; + break; + case MACHINE: + // If token contains a protocol: Check it first, and strip it away if + // it matches. If it does not match, ignore this stanza. + // If there is no protocol, only allow https protocols. + protocolSpecified = token.find("://") != std::string::npos; + if (protocolSpecified) + { + if (not APT::String::Startswith(token, Uri.Access + "://")) + { + active_token = NO; + break; + } + token.erase(0, Uri.Access.length() + 3); + } + + if (token.find('/') == std::string::npos) + { + if (Uri.Port != 0 && Uri.Host == token) + active_token = GOOD_MACHINE; + else if (lookfor.compare(0, lookfor.length() - Uri.Path.length(), token) == 0) + active_token = GOOD_MACHINE; + else + active_token = NO; + } + else + { + if (APT::String::Startswith(lookfor, token)) + active_token = GOOD_MACHINE; + else + active_token = NO; + } + + if (active_token == GOOD_MACHINE && not protocolSpecified) + { + if (Uri.Access != "https" && Uri.Access != "tor+https") + { + _error->Warning(_("%s: Credentials for %s match, but the protocol is not encrypted. Annotate with %s:// to use."), NetRCFile.Name().c_str(), token.c_str(), Uri.Access.c_str()); + active_token = NO; + } + } + break; + case GOOD_MACHINE: + if (token == "login") + active_token = LOGIN; + else if (token == "password") + active_token = PASSWORD; + else if (token == "machine") + { + if (Debug) + std::clog << "MaybeAddAuth: Found matching host adding '" << Uri.User << "' and '" << Uri.Password << "' for " + << (std::string)Uri << " from " << NetRCFile.Name() << std::endl; + return true; + } + break; + case LOGIN: + std::swap(Uri.User, token); + active_token = GOOD_MACHINE; + break; + case PASSWORD: + std::swap(Uri.Password, token); + active_token = GOOD_MACHINE; + break; + } + } + if (active_token == GOOD_MACHINE) + { + if (Debug) + std::clog << "MaybeAddAuth: Found matching host adding '" << Uri.User << "' and '" << Uri.Password << "' for " + << (std::string)Uri << " from " << NetRCFile.Name() << std::endl; + return true; + } + else if (active_token == NO) + { + if (Debug) + std::clog << "MaybeAddAuth: Found no matching host for " + << (std::string)Uri << " from " << NetRCFile.Name() << std::endl; + return true; + } + else if (Debug) + { + std::clog << "MaybeAddAuth: Found no matching host (syntax error: token:"; + switch (active_token) + { + case NO: std::clog << "NO"; break; + case MACHINE: std::clog << "MACHINE"; break; + case GOOD_MACHINE: std::clog << "GOOD_MACHINE"; break; + case LOGIN: std::clog << "LOGIN"; break; + case PASSWORD: std::clog << "PASSWORD"; break; + } + std::clog << ") for " << (std::string)Uri << " from " << NetRCFile.Name() << std::endl; + } + return false; +} diff --git a/apt-pkg/contrib/netrc.h b/apt-pkg/contrib/netrc.h new file mode 100644 index 0000000..2d202d1 --- /dev/null +++ b/apt-pkg/contrib/netrc.h @@ -0,0 +1,30 @@ +// -*- mode: cpp; mode: fold -*- +// SPDX-License-Identifier: GPL-2.0+ +// Description /*{{{*/ +/* ###################################################################### + + netrc file parser - returns the login and password of a give host in + a specified netrc-type file + + This file had this historic note, but now includes further changes + under the GPL-2.0+: + + Originally written by Daniel Stenberg, <daniel@haxx.se>, et al. and + placed into the Public Domain, do with it what you will. + + ##################################################################### */ + /*}}}*/ +#ifndef NETRC_H +#define NETRC_H + +#include <string> + +#include <apt-pkg/macros.h> + + + +class URI; +class FileFd; + +APT_PUBLIC bool MaybeAddAuth(FileFd &NetRCFile, URI &Uri); +#endif diff --git a/apt-pkg/contrib/progress.cc b/apt-pkg/contrib/progress.cc new file mode 100644 index 0000000..eb688b9 --- /dev/null +++ b/apt-pkg/contrib/progress.cc @@ -0,0 +1,229 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + OpProgress - Operation Progress + + ##################################################################### */ + /*}}}*/ +// Include Files /*{{{*/ +#include <config.h> + +#include <apt-pkg/configuration.h> +#include <apt-pkg/error.h> +#include <apt-pkg/progress.h> + +#include <cmath> +#include <chrono> +#include <cstring> +#include <iostream> +#include <string> +#include <stdio.h> +#include <sys/time.h> + +#include <apti18n.h> + /*}}}*/ + +using namespace std; + +// OpProgress::OpProgress - Constructor /*{{{*/ +// --------------------------------------------------------------------- +/* */ +OpProgress::OpProgress() : Current(0), Total(0), Size(0), SubTotal(1), + LastPercent(0), Percent(0) +{ + memset(&LastTime,0,sizeof(LastTime)); +} + /*}}}*/ +// OpProgress::Progress - Sub progress with no state change /*{{{*/ +// --------------------------------------------------------------------- +/* Current is the Base Overall progress in units of Total. Cur is the sub + progress in units of SubTotal. Size is a scaling factor that says what + percent of Total SubTotal is. */ +void OpProgress::Progress(unsigned long long Cur) +{ + if (Total == 0 || Size == 0 || SubTotal == 0) + Percent = 0; + else + Percent = (Current + Cur/((double)SubTotal)*Size)*100.0/Total; + Update(); +} + /*}}}*/ +// OpProgress::OverallProgress - Set the overall progress /*{{{*/ +// --------------------------------------------------------------------- +/* */ +void OpProgress::OverallProgress(unsigned long long Current, unsigned long long Total, + unsigned long long Size,const string &Op) +{ + this->Current = Current; + this->Total = Total; + this->Size = Size; + this->Op = Op; + SubOp = string(); + if (Total == 0) + Percent = 0; + else + Percent = Current*100.0/Total; + Update(); +} + /*}}}*/ +// OpProgress::SubProgress - Set the sub progress state /*{{{*/ +// --------------------------------------------------------------------- +/* */ +void OpProgress::SubProgress(unsigned long long SubTotal,const string &Op, + float const Percent) +{ + this->SubTotal = SubTotal; + if (Op.empty() == false) + SubOp = Op; + if (Total == 0 || Percent == 0) + this->Percent = 0; + else if (Percent != -1) + this->Percent = this->Current += (Size*Percent)/SubTotal; + else + this->Percent = Current*100.0/Total; + Update(); +} + /*}}}*/ +// OpProgress::CheckChange - See if the display should be updated /*{{{*/ +// --------------------------------------------------------------------- +/* Progress calls are made so frequently that if every one resulted in + an update the display would be swamped and the system much slower. + This provides an upper bound on the update rate. */ +bool OpProgress::CheckChange(float Interval) +{ + // For absolute progress, we assume every call is relevant. + if (_config->FindB("APT::Internal::OpProgress::Absolute", false)) + return true; + // New major progress indication + if (Op != LastOp) + { + MajorChange = true; + LastOp = Op; + return true; + } + MajorChange = false; + + if (SubOp != LastSubOp) + { + LastSubOp = SubOp; + return true; + } + + if (std::lround(LastPercent) == std::lround(Percent)) + return false; + + LastPercent = Percent; + + if (Interval == 0) + return false; + + // Check time delta + auto const Now = std::chrono::steady_clock::now().time_since_epoch(); + auto const Now_sec = std::chrono::duration_cast<std::chrono::seconds>(Now); + auto const Now_usec = std::chrono::duration_cast<std::chrono::microseconds>(Now - Now_sec); + struct timeval NowTime = { static_cast<time_t>(Now_sec.count()), static_cast<suseconds_t>(Now_usec.count()) }; + + std::chrono::duration<decltype(Interval)> Delta = + std::chrono::seconds(NowTime.tv_sec - LastTime.tv_sec) + + std::chrono::microseconds(NowTime.tv_usec - LastTime.tv_usec); + + if (Delta.count() < Interval) + return false; + LastTime = NowTime; + return true; +} + /*}}}*/ +// OpTextProgress::OpTextProgress - Constructor /*{{{*/ +// --------------------------------------------------------------------- +/* */ +OpTextProgress::OpTextProgress(Configuration &Config) : + NoUpdate(false), NoDisplay(false), LastLen(0) +{ + if (Config.FindI("quiet",0) >= 1 || Config.FindB("quiet::NoUpdate", false) == true) + NoUpdate = true; + if (Config.FindI("quiet",0) >= 2 || Config.FindB("quiet::NoProgress", false) == true) + NoDisplay = true; +} + /*}}}*/ +// OpTextProgress::Done - Clean up the display /*{{{*/ +// --------------------------------------------------------------------- +/* */ +void OpTextProgress::Done() +{ + if (NoUpdate == false && OldOp.empty() == false) + { + char S[300]; + if (_error->PendingError() == true) + snprintf(S,sizeof(S),_("%c%s... Error!"),'\r',OldOp.c_str()); + else + snprintf(S,sizeof(S),_("%c%s... Done"),'\r',OldOp.c_str()); + Write(S); + cout << endl; + OldOp = string(); + } + + if (NoUpdate == true && NoDisplay == false && OldOp.empty() == false) + { + OldOp = string(); + cout << endl; + } +} + /*}}}*/ +// OpTextProgress::Update - Simple text spinner /*{{{*/ +// --------------------------------------------------------------------- +/* */ +void OpTextProgress::Update() +{ + if (CheckChange((NoUpdate == true?0:0.7)) == false) + return; + + // No percent spinner + if (NoUpdate == true) + { + if (MajorChange == false) + return; + if (NoDisplay == false) + { + if (OldOp.empty() == false) + cout << endl; + OldOp = "a"; + cout << Op << _("...") << flush; + } + + return; + } + + // Erase the old text and 'log' the event + char S[300]; + if (MajorChange == true && OldOp.empty() == false) + { + snprintf(S,sizeof(S),"\r%s",OldOp.c_str()); + Write(S); + cout << endl; + } + + // Print the spinner. Absolute progress shows us a time progress. + if (_config->FindB("APT::Internal::OpProgress::Absolute", false) && Total != -1llu) + snprintf(S, sizeof(S), _("%c%s... %llu/%llus"), '\r', Op.c_str(), Current, Total); + else if (_config->FindB("APT::Internal::OpProgress::Absolute", false)) + snprintf(S, sizeof(S), _("%c%s... %llus"), '\r', Op.c_str(), Current); + else + snprintf(S, sizeof(S), _("%c%s... %u%%"), '\r', Op.c_str(), (unsigned int)Percent); + Write(S); + + OldOp = Op; +} + /*}}}*/ +// OpTextProgress::Write - Write the progress string /*{{{*/ +// --------------------------------------------------------------------- +/* This space fills the end to overwrite the previous text */ +void OpTextProgress::Write(const char *S) +{ + cout << S; + for (unsigned int I = strlen(S); I < LastLen; I++) + cout << ' '; + cout << '\r' << flush; + LastLen = strlen(S); +} + /*}}}*/ diff --git a/apt-pkg/contrib/progress.h b/apt-pkg/contrib/progress.h new file mode 100644 index 0000000..d6a698a --- /dev/null +++ b/apt-pkg/contrib/progress.h @@ -0,0 +1,87 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + OpProgress - Operation Progress + + This class allows lengthy operations to communicate their progress + to the GUI. The progress model is simple and is not designed to handle + the complex case of the multi-activity acquire class. + + The model is based on the concept of an overall operation consisting + of a series of small sub operations. Each sub operation has it's own + completion status and the overall operation has it's completion status. + The units of the two are not mixed and are completely independent. + + The UI is expected to subclass this to provide the visuals to the user. + + ##################################################################### */ + /*}}}*/ +#ifndef PKGLIB_PROGRESS_H +#define PKGLIB_PROGRESS_H + +#include <apt-pkg/macros.h> +#include <string> +#include <sys/time.h> + + +class Configuration; +class APT_PUBLIC OpProgress +{ + friend class OpTextProgress; + unsigned long long Current; + unsigned long long Total; + unsigned long long Size; + unsigned long long SubTotal; + float LastPercent; + + // Change reduction code + struct timeval LastTime; + std::string LastOp; + std::string LastSubOp; + + protected: + + std::string Op; + std::string SubOp; + float Percent; + + bool MajorChange; + + bool CheckChange(float Interval = 0.7); + virtual void Update() {}; + + public: + + void Progress(unsigned long long Current); + void SubProgress(unsigned long long SubTotal, const std::string &Op = "", float const Percent = -1); + void OverallProgress(unsigned long long Current,unsigned long long Total, + unsigned long long Size,const std::string &Op); + virtual void Done() {}; + + OpProgress(); + virtual ~OpProgress() {}; +}; + +class APT_PUBLIC OpTextProgress : public OpProgress +{ + protected: + + std::string OldOp; + bool NoUpdate; + bool NoDisplay; + unsigned long LastLen; + virtual void Update() APT_OVERRIDE; + void Write(const char *S); + + public: + + virtual void Done() APT_OVERRIDE; + + explicit OpTextProgress(bool NoUpdate = false) : NoUpdate(NoUpdate), + NoDisplay(false), LastLen(0) {}; + explicit OpTextProgress(Configuration &Config); + virtual ~OpTextProgress() {Done();}; +}; + +#endif diff --git a/apt-pkg/contrib/proxy.cc b/apt-pkg/contrib/proxy.cc new file mode 100644 index 0000000..a99f44f --- /dev/null +++ b/apt-pkg/contrib/proxy.cc @@ -0,0 +1,99 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + Proxy - Proxy related functions + + ##################################################################### */ + /*}}}*/ +// Include Files /*{{{*/ +#include <config.h> + +#include <apt-pkg/configuration.h> +#include <apt-pkg/error.h> +#include <apt-pkg/fileutl.h> +#include <apt-pkg/strutl.h> + +#include <algorithm> +#include <iostream> +#include <fcntl.h> +#include <unistd.h> + +#include "proxy.h" + /*}}}*/ + +// AutoDetectProxy - auto detect proxy /*{{{*/ +// --------------------------------------------------------------------- +/* */ +static std::vector<std::string> CompatibleProxies(URI const &URL) +{ + if (URL.Access == "http" || URL.Access == "https") + return {"http", "https", "socks5h"}; + return {URL.Access}; +} + +bool AutoDetectProxy(URI &URL) +{ + // we support both http/https debug options + bool Debug = _config->FindB("Debug::Acquire::"+URL.Access,false); + + // the user already explicitly set a proxy for this host + if(_config->Find("Acquire::"+URL.Access+"::proxy::"+URL.Host, "") != "") + return true; + + // option is "Acquire::http::Proxy-Auto-Detect" but we allow the old + // name without the dash ("-") + std::string AutoDetectProxyCmd = _config->Find("Acquire::"+URL.Access+"::Proxy-Auto-Detect", + _config->Find("Acquire::"+URL.Access+"::ProxyAutoDetect")); + + if (AutoDetectProxyCmd.empty()) + return true; + + if (Debug) + std::clog << "Using auto proxy detect command: " << AutoDetectProxyCmd << std::endl; + + if (faccessat(AT_FDCWD, AutoDetectProxyCmd.c_str(), R_OK | X_OK, AT_EACCESS) != 0) + return _error->Errno("access", "ProxyAutoDetect command '%s' can not be executed!", AutoDetectProxyCmd.c_str()); + + std::string const urlstring = URL; + std::vector<const char *> Args; + Args.push_back(AutoDetectProxyCmd.c_str()); + Args.push_back(urlstring.c_str()); + Args.push_back(nullptr); + FileFd PipeFd; + pid_t Child; + if (Popen(&Args[0], PipeFd, Child, FileFd::ReadOnly, false, true) == false) + return _error->Error("ProxyAutoDetect command '%s' failed!", AutoDetectProxyCmd.c_str()); + char buf[512]; + bool const goodread = PipeFd.ReadLine(buf, sizeof(buf)) != nullptr; + PipeFd.Close(); + if (ExecWait(Child, "ProxyAutoDetect") == false) + return false; + // no output means the detector has no idea which proxy to use + // and apt will use the generic proxy settings + if (goodread == false) + return true; + auto const cleanedbuf = _strstrip(buf); + // We warn about this as the implementor probably meant to use DIRECT instead + if (cleanedbuf[0] == '\0') + { + _error->Warning("ProxyAutoDetect command returned an empty line"); + return true; + } + + if (Debug) + std::clog << "auto detect command returned: '" << cleanedbuf << "'" << std::endl; + + auto compatibleTypes = CompatibleProxies(URL); + bool compatible = strcmp(cleanedbuf, "DIRECT") == 0 || + compatibleTypes.end() != std::find_if(compatibleTypes.begin(), + compatibleTypes.end(), [cleanedbuf](std::string &compat) { + return strstr(cleanedbuf, compat.c_str()) == cleanedbuf; + }); + + if (compatible) + _config->Set("Acquire::"+URL.Access+"::proxy::"+URL.Host, cleanedbuf); + + return true; +} + /*}}}*/ diff --git a/apt-pkg/contrib/proxy.h b/apt-pkg/contrib/proxy.h new file mode 100644 index 0000000..f6d70ea --- /dev/null +++ b/apt-pkg/contrib/proxy.h @@ -0,0 +1,16 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + Proxy - Proxy operations + + ##################################################################### */ + /*}}}*/ +#ifndef PKGLIB_PROXY_H +#define PKGLIB_PROXY_H + +class URI; +APT_PUBLIC bool AutoDetectProxy(URI &URL); + + +#endif diff --git a/apt-pkg/contrib/srvrec.cc b/apt-pkg/contrib/srvrec.cc new file mode 100644 index 0000000..3eb5f1d --- /dev/null +++ b/apt-pkg/contrib/srvrec.cc @@ -0,0 +1,211 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + SRV record support + + ##################################################################### */ + /*}}}*/ +#include <config.h> + +#include <netdb.h> + +#include <arpa/nameser.h> +#include <arpa/inet.h> +#include <netinet/in.h> +#include <resolv.h> +#include <time.h> + +#include <algorithm> +#include <memory> +#include <tuple> + +#include <apt-pkg/configuration.h> +#include <apt-pkg/error.h> +#include <apt-pkg/strutl.h> + +#include "srvrec.h" + +bool SrvRec::operator==(SrvRec const &other) const +{ + return (std::tie(target, priority, weight, port) == + std::tie(other.target, other.priority, other.weight, other.port)); +} + +bool GetSrvRecords(std::string host, int port, std::vector<SrvRec> &Result) +{ + // try SRV only for hostnames, not for IP addresses + { + struct in_addr addr4; + struct in6_addr addr6; + if (inet_pton(AF_INET, host.c_str(), &addr4) == 1 || + inet_pton(AF_INET6, host.c_str(), &addr6) == 1) + return true; + } + + std::string target; + int res; + struct servent s_ent_buf; + struct servent *s_ent = nullptr; + std::vector<char> buf(1024); + + res = getservbyport_r(htons(port), "tcp", &s_ent_buf, buf.data(), buf.size(), &s_ent); + if (res != 0 || s_ent == nullptr) + return false; + + strprintf(target, "_%s._tcp.%s", s_ent->s_name, host.c_str()); + return GetSrvRecords(target, Result); +} + +bool GetSrvRecords(std::string name, std::vector<SrvRec> &Result) +{ + unsigned char answer[PACKETSZ]; + int answer_len, compressed_name_len; + int answer_count; +#if __RES >= 19991006 + struct __res_state res; + + if (res_ninit(&res) != 0) + return _error->Errno("res_init", "Failed to init resolver"); + + // Close on return + std::shared_ptr<void> guard(&res, res_nclose); + + answer_len = res_nquery(&res, name.c_str(), C_IN, T_SRV, answer, sizeof(answer)); +#else + if (res_init() != 0) + return _error->Errno("res_init", "Failed to init resolver"); + + answer_len = res_query(name.c_str(), C_IN, T_SRV, answer, sizeof(answer)); +#endif //__RES >= 19991006 + if (answer_len == -1) + return false; + if (answer_len < (int)sizeof(HEADER)) + return _error->Warning("Not enough data from res_query (%i)", answer_len); + + // check the header + HEADER *header = (HEADER*)answer; + if (header->rcode != NOERROR) + return _error->Warning("res_query returned rcode %i", header->rcode); + answer_count = ntohs(header->ancount); + if (answer_count <= 0) + return _error->Warning("res_query returned no answers (%i) ", answer_count); + + // skip the header + compressed_name_len = dn_skipname(answer+sizeof(HEADER), answer+answer_len); + if(compressed_name_len < 0) + return _error->Warning("dn_skipname failed %i", compressed_name_len); + + // pt points to the first answer record, go over all of them now + unsigned char *pt = answer+sizeof(HEADER)+compressed_name_len+QFIXEDSZ; + while ((int)Result.size() < answer_count && pt < answer+answer_len) + { + u_int16_t type, klass, priority, weight, port, dlen; + char buf[MAXDNAME]; + + compressed_name_len = dn_skipname(pt, answer+answer_len); + if (compressed_name_len < 0) + return _error->Warning("dn_skipname failed (2): %i", + compressed_name_len); + pt += compressed_name_len; + if (((answer+answer_len) - pt) < 16) + return _error->Warning("packet too short"); + + // extract the data out of the result buffer + #define extract_u16(target, p) target = *p++ << 8; target |= *p++; + + extract_u16(type, pt); + if(type != T_SRV) + return _error->Warning("Unexpected type excepted %x != %x", + T_SRV, type); + extract_u16(klass, pt); + if(klass != C_IN) + return _error->Warning("Unexpected class excepted %x != %x", + C_IN, klass); + pt += 4; // ttl + extract_u16(dlen, pt); + extract_u16(priority, pt); + extract_u16(weight, pt); + extract_u16(port, pt); + + #undef extract_u16 + + compressed_name_len = dn_expand(answer, answer+answer_len, pt, buf, sizeof(buf)); + if(compressed_name_len < 0) + return _error->Warning("dn_expand failed %i", compressed_name_len); + pt += compressed_name_len; + + // add it to our class + Result.emplace_back(buf, priority, weight, port); + } + + // implement load balancing as specified in RFC-2782 + + // sort them by priority + std::stable_sort(Result.begin(), Result.end()); + + if (_config->FindB("Debug::Acquire::SrvRecs", false)) + for(auto const &R : Result) + std::cerr << "SrvRecs: got " << R.target + << " prio: " << R.priority + << " weight: " << R.weight + << '\n'; + + return true; +} + +SrvRec PopFromSrvRecs(std::vector<SrvRec> &Recs) +{ + // FIXME: instead of the simplistic shuffle below use the algorithm + // described in rfc2782 (with weights) + // and figure out how the weights need to be adjusted if + // a host refuses connections + +#if 0 // all code below is only needed for the weight adjusted selection + // assign random number ranges + int prev_weight = 0; + int prev_priority = 0; + for(std::vector<SrvRec>::iterator I = Result.begin(); + I != Result.end(); ++I) + { + if(prev_priority != I->priority) + prev_weight = 0; + I->random_number_range_start = prev_weight; + I->random_number_range_end = prev_weight + I->weight; + prev_weight = I->random_number_range_end; + prev_priority = I->priority; + + if (_config->FindB("Debug::Acquire::SrvRecs", false) == true) + std::cerr << "SrvRecs: got " << I->target + << " prio: " << I->priority + << " weight: " << I->weight + << std::endl; + } + + // go over the code in reverse order and note the max random range + int max = 0; + prev_priority = 0; + for(std::vector<SrvRec>::iterator I = Result.end(); + I != Result.begin(); --I) + { + if(prev_priority != I->priority) + max = I->random_number_range_end; + I->random_number_range_max = max; + } +#endif + + // shuffle in a very simplistic way for now (equal weights) + std::vector<SrvRec>::iterator I = Recs.begin(); + std::vector<SrvRec>::iterator const J = std::find_if(Recs.begin(), Recs.end(), + [&I](SrvRec const &J) { return I->priority != J.priority; }); + + // clock seems random enough. + I += std::max(static_cast<clock_t>(0), clock()) % std::distance(I, J); + SrvRec const selected = std::move(*I); + Recs.erase(I); + + if (_config->FindB("Debug::Acquire::SrvRecs", false) == true) + std::cerr << "PopFromSrvRecs: selecting " << selected.target << std::endl; + + return selected; +} diff --git a/apt-pkg/contrib/srvrec.h b/apt-pkg/contrib/srvrec.h new file mode 100644 index 0000000..1e981d3 --- /dev/null +++ b/apt-pkg/contrib/srvrec.h @@ -0,0 +1,57 @@ +// -*- mode: cpp; mode: fold -*- +// Description /*{{{*/ +/* ###################################################################### + + SRV record support + + ##################################################################### */ + /*}}}*/ +#ifndef SRVREC_H +#define SRVREC_H + +#include <string> +#include <vector> +#include <arpa/nameser.h> +#include <sys/types.h> + +#include <apt-pkg/macros.h> + +class APT_PUBLIC SrvRec +{ + public: + std::string target; + u_int16_t priority; + u_int16_t weight; + u_int16_t port; + + // each server is assigned a interval [start, end] in the space of [0, max] + int random_number_range_start; + int random_number_range_end; + int random_number_range_max; + + bool operator<(SrvRec const &other) const { + return this->priority < other.priority; + } + bool operator==(SrvRec const &other) const; + + SrvRec(std::string const Target, u_int16_t const Priority, + u_int16_t const Weight, u_int16_t const Port) : + target(Target), priority(Priority), weight(Weight), port(Port), + random_number_range_start(0), random_number_range_end(0), + random_number_range_max(0) {} +}; + +/** \brief Get SRV records from host/port (builds the query string internally) + */ +APT_PUBLIC bool GetSrvRecords(std::string name, std::vector<SrvRec> &Result); + +/** \brief Get SRV records for query string like: _http._tcp.example.com + */ +APT_PUBLIC bool GetSrvRecords(std::string host, int port, std::vector<SrvRec> &Result); + +/** \brief Pop a single SRV record from the vector of SrvRec taking + * priority and weight into account + */ +APT_PUBLIC SrvRec PopFromSrvRecs(std::vector<SrvRec> &Recs); + +#endif diff --git a/apt-pkg/contrib/string_view.h b/apt-pkg/contrib/string_view.h new file mode 100644 index 0000000..04f6ff1 --- /dev/null +++ b/apt-pkg/contrib/string_view.h @@ -0,0 +1,163 @@ +/* + * Basic implementation of string_view + * + * (C) 2015 Julian Andres Klode <jak@debian.org> + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + */ + +#if !defined(APT_STRINGVIEW_H) +#define APT_STRINGVIEW_H +#include <apt-pkg/macros.h> +#include <string> +#include <string.h> + +namespace APT { + +/** + * \brief Simple subset of std::string_view from C++17 + * + * This is an internal implementation of the subset of std::string_view + * used by APT. It is not meant to be used in programs, only inside the + * library for performance critical paths. + */ +class StringView { + const char *data_; + size_t size_; + +public: + static constexpr size_t npos = static_cast<size_t>(-1); + static_assert(APT::StringView::npos == std::string::npos, "npos values are different"); + + /* Constructors */ + constexpr StringView() : data_(""), size_(0) {} + constexpr StringView(const char *data, size_t size) : data_(data), size_(size) {} + + StringView(const char *data) : data_(data), size_(strlen(data)) {} + StringView(std::string const & str): data_(str.data()), size_(str.size()) {} + + /* Modifiers */ + void remove_prefix(size_t n) { data_ += n; size_ -= n; } + void remove_suffix(size_t n) { size_ -= n; } + void clear() { size_ = 0; } + + /* Viewers */ + constexpr StringView substr(size_t pos, size_t n = npos) const { + return StringView(data_ + pos, n > (size_ - pos) ? (size_ - pos) : n); + } + + size_t find(int c, size_t pos) const { + if (pos == 0) + return find(c); + size_t const found = substr(pos).find(c); + if (found == npos) + return npos; + return pos + found; + } + size_t find(int c) const { + const char *found = static_cast<const char*>(memchr(data_, c, size_)); + + if (found == NULL) + return npos; + + return found - data_; + } + + size_t rfind(int c, size_t pos) const { + if (pos == npos) + return rfind(c); + return APT::StringView(data_, pos).rfind(c); + } + size_t rfind(int c) const { + const char *found = static_cast<const char*>(memrchr(data_, c, size_)); + + if (found == NULL) + return npos; + + return found - data_; + } + + size_t find(APT::StringView const needle) const { + if (needle.empty()) + return npos; + if (needle.length() == 1) + return find(*needle.data()); + size_t found = 0; + while ((found = find(*needle.data(), found)) != npos) { + if (compare(found, needle.length(), needle) == 0) + return found; + ++found; + } + return found; + } + size_t find(APT::StringView const needle, size_t pos) const { + if (pos == 0) + return find(needle); + size_t const found = substr(pos).find(needle); + if (found == npos) + return npos; + return pos + found; + } + + /* Conversions */ + std::string to_string() const { + return std::string(data_, size_); + } + + /* Comparisons */ + int compare(size_t pos, size_t n, StringView other) const { + return substr(pos, n).compare(other); + } + + int compare(StringView other) const { + int res; + + res = memcmp(data_, other.data_, std::min(size_, other.size_)); + if (res != 0) + return res; + if (size_ == other.size_) + return res; + + return (size_ > other.size_) ? 1 : -1; + } + + /* Optimization: If size not equal, string cannot be equal */ + bool operator ==(StringView other) const { return size_ == other.size_ && compare(other) == 0; } + bool operator !=(StringView other) const { return !(*this == other); } + + /* Accessors */ + constexpr bool empty() const { return size_ == 0; } + constexpr const char* data() const { return data_; } + constexpr const char* begin() const { return data_; } + constexpr const char* end() const { return data_ + size_; } + constexpr char operator [](size_t i) const { return data_[i]; } + constexpr size_t size() const { return size_; } + constexpr size_t length() const { return size_; } +}; + +/** + * \brief Faster comparison for string views (compare size before data) + * + * Still stable, but faster than the normal ordering. */ +static inline int StringViewCompareFast(StringView a, StringView b) { + if (a.size() != b.size()) + return a.size() - b.size(); + + return memcmp(a.data(), b.data(), a.size()); +} + +static constexpr inline APT::StringView operator""_sv(const char *data, size_t size) +{ + return APT::StringView(data, size); +} +} + +inline bool operator ==(const char *other, APT::StringView that); +inline bool operator ==(const char *other, APT::StringView that) { return that.operator==(other); } +inline bool operator ==(std::string const &other, APT::StringView that); +inline bool operator ==(std::string const &other, APT::StringView that) { return that.operator==(other); } + +#endif diff --git a/apt-pkg/contrib/strutl.cc b/apt-pkg/contrib/strutl.cc new file mode 100644 index 0000000..67100f1 --- /dev/null +++ b/apt-pkg/contrib/strutl.cc @@ -0,0 +1,1820 @@ +// -*- mode: cpp; mode: fold -*- +// SPDX-License-Identifier: GPL-2.0+ +// Description /*{{{*/ +/* ###################################################################### + + String Util - Some useful string functions. + + These have been collected from here and there to do all sorts of useful + things to strings. They are useful in file parsers, URI handlers and + especially in APT methods. + + This file had this historic note, but now includes further changes + under the GPL-2.0+: + + This source is placed in the Public Domain, do with it what you will + It was originally written by Jason Gunthorpe <jgg@gpu.srv.ualberta.ca> + + ##################################################################### */ + /*}}}*/ +// Includes /*{{{*/ +#include <config.h> + +#include <apt-pkg/error.h> +#include <apt-pkg/fileutl.h> +#include <apt-pkg/strutl.h> + +#include <algorithm> +#include <array> +#include <iomanip> +#include <limits> +#include <locale> +#include <sstream> +#include <memory> +#include <sstream> +#include <string> +#include <vector> + +#include <ctype.h> +#include <errno.h> +#include <iconv.h> +#include <regex.h> +#include <stdarg.h> +#include <stddef.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <time.h> +#include <unistd.h> +#include <wchar.h> + +#include <apti18n.h> + /*}}}*/ +using namespace std; + +// Strip - Remove white space from the front and back of a string /*{{{*/ +// --------------------------------------------------------------------- +namespace APT { + namespace String { +std::string Strip(const std::string &str) +{ + // ensure we have at least one character + if (str.empty() == true) + return str; + + char const * const s = str.c_str(); + size_t start = 0; + for (; isspace(s[start]) != 0; ++start) + ; // find the first not-space + + // string contains only whitespaces + if (s[start] == '\0') + return ""; + + size_t end = str.length() - 1; + for (; isspace(s[end]) != 0; --end) + ; // find the last not-space + + return str.substr(start, end - start + 1); +} + +bool Endswith(const std::string &s, const std::string &end) +{ + if (end.size() > s.size()) + return false; + return (s.compare(s.size() - end.size(), end.size(), end) == 0); +} + +bool Startswith(const std::string &s, const std::string &start) +{ + if (start.size() > s.size()) + return false; + return (s.compare(0, start.size(), start) == 0); +} + +std::string Join(std::vector<std::string> list, const std::string &sep) +{ + std::ostringstream oss; + for (auto it = list.begin(); it != list.end(); it++) + { + if (it != list.begin()) oss << sep; + oss << *it; + } + return oss.str(); +} + +// Returns string display length honoring multi-byte characters +size_t DisplayLength(StringView str) +{ + size_t len = 0; + + const char *p = str.data(); + const char *const end = str.end(); + + mbstate_t state{}; + while (p < end) + { + wchar_t wch; + size_t res = mbrtowc(&wch, p, end - p, &state); + switch (res) + { + case 0: + // Null wide character (i.e. L'\0') - stop + p = end; + break; + + case static_cast<size_t>(-1): + // Byte sequence is invalid. Assume that it's + // a single-byte single-width character. + len += 1; + p += 1; + + // state is undefined in this case - reset it + state = {}; + + break; + + case static_cast<size_t>(-2): + // Byte sequence is too short. Assume that it's + // an incomplete single-width character and stop. + len += 1; + p = end; + break; + + default: + len += wcwidth(wch); + p += res; + } + } + + return len; +} + +} +} + /*}}}*/ +// UTF8ToCodeset - Convert some UTF-8 string for some codeset /*{{{*/ +// --------------------------------------------------------------------- +/* This is handy to use before display some information for enduser */ +bool UTF8ToCodeset(const char *codeset, const string &orig, string *dest) +{ + iconv_t cd; + const char *inbuf; + char *inptr, *outbuf; + size_t insize, bufsize; + dest->clear(); + + cd = iconv_open(codeset, "UTF-8"); + if (cd == (iconv_t)(-1)) { + // Something went wrong + if (errno == EINVAL) + _error->Error("conversion from 'UTF-8' to '%s' not available", + codeset); + else + perror("iconv_open"); + + return false; + } + + insize = bufsize = orig.size(); + inbuf = orig.data(); + inptr = (char *)inbuf; + outbuf = new char[bufsize]; + size_t lastError = -1; + + while (insize != 0) + { + char *outptr = outbuf; + size_t outsize = bufsize; + size_t const err = iconv(cd, &inptr, &insize, &outptr, &outsize); + dest->append(outbuf, outptr - outbuf); + if (err == (size_t)(-1)) + { + switch (errno) + { + case EILSEQ: + insize--; + inptr++; + // replace a series of unknown multibytes with a single "?" + if (lastError != insize) { + lastError = insize - 1; + dest->append("?"); + } + break; + case EINVAL: + insize = 0; + break; + case E2BIG: + if (outptr == outbuf) + { + bufsize *= 2; + delete[] outbuf; + outbuf = new char[bufsize]; + } + break; + } + } + } + + delete[] outbuf; + + iconv_close(cd); + + return true; +} + /*}}}*/ +// strstrip - Remove white space from the front and back of a string /*{{{*/ +// --------------------------------------------------------------------- +/* This is handy to use when parsing a file. It also removes \n's left + over from fgets and company */ +char *_strstrip(char *String) +{ + for (;*String != 0 && (*String == ' ' || *String == '\t'); String++); + + if (*String == 0) + return String; + return _strrstrip(String); +} + /*}}}*/ +// strrstrip - Remove white space from the back of a string /*{{{*/ +// --------------------------------------------------------------------- +char *_strrstrip(char *String) +{ + char *End = String + strlen(String) - 1; + for (;End != String - 1 && (*End == ' ' || *End == '\t' || *End == '\n' || + *End == '\r'); End--); + End++; + *End = 0; + return String; +} + /*}}}*/ +// strtabexpand - Converts tabs into 8 spaces /*{{{*/ +// --------------------------------------------------------------------- +/* */ +char *_strtabexpand(char *String,size_t Len) +{ + for (char *I = String; I != I + Len && *I != 0; I++) + { + if (*I != '\t') + continue; + if (I + 8 > String + Len) + { + *I = 0; + return String; + } + + /* Assume the start of the string is 0 and find the next 8 char + division */ + int Len; + if (String == I) + Len = 1; + else + Len = 8 - ((String - I) % 8); + Len -= 2; + if (Len <= 0) + { + *I = ' '; + continue; + } + + memmove(I + Len,I + 1,strlen(I) + 1); + for (char *J = I; J + Len != I; *I = ' ', I++); + } + return String; +} + /*}}}*/ +// ParseQuoteWord - Parse a single word out of a string /*{{{*/ +// --------------------------------------------------------------------- +/* This grabs a single word, converts any % escaped characters to their + proper values and advances the pointer. Double quotes are understood + and striped out as well. This is for URI/URL parsing. It also can + understand [] brackets.*/ +bool ParseQuoteWord(const char *&String,string &Res) +{ + // Skip leading whitespace + const char *C = String; + for (; *C == ' '; C++) + ; + if (*C == 0) + return false; + + // Jump to the next word + for (;*C != 0 && isspace(*C) == 0; C++) + { + if (*C == '"') + { + C = strchr(C + 1, '"'); + if (C == NULL) + return false; + } + if (*C == '[') + { + C = strchr(C + 1, ']'); + if (C == NULL) + return false; + } + } + + // Now de-quote characters + Res.clear(); + Res.reserve(C - String); + char Tmp[3]; + const char *Start = String; + while (Start != C) + { + if (*Start == '%' && Start + 2 < C && + isxdigit(Start[1]) && isxdigit(Start[2])) + { + Tmp[0] = Start[1]; + Tmp[1] = Start[2]; + Tmp[2] = 0; + Res.push_back(static_cast<char>(strtol(Tmp, 0, 16))); + Start += 3; + continue; + } + if (*Start != '"') + Res.push_back(*Start); + ++Start; + } + + // Skip ending white space + for (; isspace(*C) != 0; C++) + ; + String = C; + return true; +} + /*}}}*/ +// ParseCWord - Parses a string like a C "" expression /*{{{*/ +// --------------------------------------------------------------------- +/* This expects a series of space separated strings enclosed in ""'s. + It concatenates the ""'s into a single string. */ +bool ParseCWord(const char *&String,string &Res) +{ + // Skip leading whitespace + const char *C = String; + for (; *C == ' '; C++) + ; + if (*C == 0) + return false; + + Res.clear(); + Res.reserve(strlen(String)); + for (; *C != 0; ++C) + { + if (*C == '"') + { + for (C++; *C != 0 && *C != '"'; C++) + Res.push_back(*C); + + if (*C == 0) + return false; + + continue; + } + + if (C != String && isspace(*C) != 0 && isspace(C[-1]) != 0) + continue; + if (isspace(*C) == 0) + return false; + Res.push_back(' '); + } + String = C; + return true; +} + /*}}}*/ +// QuoteString - Convert a string into quoted from /*{{{*/ +// --------------------------------------------------------------------- +/* */ +string QuoteString(const string &Str, const char *Bad) +{ + std::stringstream Res; + for (string::const_iterator I = Str.begin(); I != Str.end(); ++I) + { + if (strchr(Bad,*I) != 0 || isprint(*I) == 0 || + *I == 0x25 || // percent '%' char + *I <= 0x20 || *I >= 0x7F) // control chars + { + ioprintf(Res, "%%%02hhx", *I); + } + else + Res << *I; + } + return Res.str(); +} + /*}}}*/ +// DeQuoteString - Convert a string from quoted from /*{{{*/ +// --------------------------------------------------------------------- +/* This undoes QuoteString */ +string DeQuoteString(const string &Str) +{ + return DeQuoteString(Str.begin(),Str.end()); +} +string DeQuoteString(string::const_iterator const &begin, + string::const_iterator const &end) +{ + string Res; + for (string::const_iterator I = begin; I != end; ++I) + { + if (*I == '%' && I + 2 < end && + isxdigit(I[1]) && isxdigit(I[2])) + { + char Tmp[3]; + Tmp[0] = I[1]; + Tmp[1] = I[2]; + Tmp[2] = 0; + Res += (char)strtol(Tmp,0,16); + I += 2; + continue; + } + else + Res += *I; + } + return Res; +} + + /*}}}*/ +// SizeToStr - Convert a long into a human readable size /*{{{*/ +// --------------------------------------------------------------------- +/* A max of 4 digits are shown before conversion to the next highest unit. + The max length of the string will be 5 chars unless the size is > 10 + YottaBytes (E24) */ +string SizeToStr(double Size) +{ + double ASize; + if (Size >= 0) + ASize = Size; + else + ASize = -1*Size; + + /* bytes, KiloBytes, MegaBytes, GigaBytes, TeraBytes, PetaBytes, + ExaBytes, ZettaBytes, YottaBytes */ + char Ext[] = {'\0','k','M','G','T','P','E','Z','Y'}; + int I = 0; + while (I <= 8) + { + if (ASize < 100 && I != 0) + { + std::string S; + strprintf(S, "%'.1f %c", ASize, Ext[I]); + return S; + } + + if (ASize < 10000) + { + std::string S; + strprintf(S, "%'.0f %c", ASize, Ext[I]); + return S; + } + ASize /= 1000.0; + I++; + } + return ""; +} + /*}}}*/ +// TimeToStr - Convert the time into a string /*{{{*/ +// --------------------------------------------------------------------- +/* Converts a number of seconds to a hms format */ +string TimeToStr(unsigned long Sec) +{ + std::string S; + if (Sec > 60*60*24) + { + //TRANSLATOR: d means days, h means hours, min means minutes, s means seconds + strprintf(S,_("%lid %lih %limin %lis"),Sec/60/60/24,(Sec/60/60) % 24,(Sec/60) % 60,Sec % 60); + } + else if (Sec > 60*60) + { + //TRANSLATOR: h means hours, min means minutes, s means seconds + strprintf(S,_("%lih %limin %lis"),Sec/60/60,(Sec/60) % 60,Sec % 60); + } + else if (Sec > 60) + { + //TRANSLATOR: min means minutes, s means seconds + strprintf(S,_("%limin %lis"),Sec/60,Sec % 60); + } + else + { + //TRANSLATOR: s means seconds + strprintf(S,_("%lis"),Sec); + } + return S; +} + /*}}}*/ +// SubstVar - Substitute a string for another string /*{{{*/ +// --------------------------------------------------------------------- +/* This replaces all occurrences of Subst with Contents in Str. */ +string SubstVar(const string &Str,const string &Subst,const string &Contents) +{ + if (Subst.empty() == true) + return Str; + + string::size_type Pos = 0; + string::size_type OldPos = 0; + string Temp; + + while (OldPos < Str.length() && + (Pos = Str.find(Subst,OldPos)) != string::npos) + { + if (OldPos != Pos) + Temp.append(Str, OldPos, Pos - OldPos); + if (Contents.empty() == false) + Temp.append(Contents); + OldPos = Pos + Subst.length(); + } + + if (OldPos == 0) + return Str; + + if (OldPos >= Str.length()) + return Temp; + + Temp.append(Str, OldPos, string::npos); + return Temp; +} +string SubstVar(string Str,const struct SubstVar *Vars) +{ + for (; Vars->Subst != 0; Vars++) + Str = SubstVar(Str,Vars->Subst,*Vars->Contents); + return Str; +} + /*}}}*/ +// OutputInDepth - return a string with separator multiplied with depth /*{{{*/ +// --------------------------------------------------------------------- +/* Returns a string with the supplied separator depth + 1 times in it */ +std::string OutputInDepth(const unsigned long Depth, const char* Separator) +{ + std::string output = ""; + for(unsigned long d=Depth+1; d > 0; d--) + output.append(Separator); + return output; +} + /*}}}*/ +// URItoFileName - Convert the uri into a unique file name /*{{{*/ +// --------------------------------------------------------------------- +/* This converts a URI into a safe filename. It quotes all unsafe characters + and converts / to _ and removes the scheme identifier. The resulting + file name should be unique and never occur again for a different file */ +string URItoFileName(const string &URI) +{ + // Nuke 'sensitive' items + ::URI U(URI); + U.User.clear(); + U.Password.clear(); + U.Access.clear(); + + // "\x00-\x20{}|\\\\^\\[\\]<>\"\x7F-\xFF"; + string NewURI = QuoteString(U,"\\|{}[]<>\"^~_=!@#$%^&*"); + replace(NewURI.begin(),NewURI.end(),'/','_'); + return NewURI; +} + /*}}}*/ +// Base64Encode - Base64 Encoding routine for short strings /*{{{*/ +// --------------------------------------------------------------------- +/* This routine performs a base64 transformation on a string. It was ripped + from wget and then patched and bug fixed. + + This spec can be found in rfc2045 */ +string Base64Encode(const string &S) +{ + // Conversion table. + static char tbl[64] = {'A','B','C','D','E','F','G','H', + 'I','J','K','L','M','N','O','P', + 'Q','R','S','T','U','V','W','X', + 'Y','Z','a','b','c','d','e','f', + 'g','h','i','j','k','l','m','n', + 'o','p','q','r','s','t','u','v', + 'w','x','y','z','0','1','2','3', + '4','5','6','7','8','9','+','/'}; + + // Pre-allocate some space + string Final; + Final.reserve((4*S.length() + 2)/3 + 2); + + /* Transform the 3x8 bits to 4x6 bits, as required by + base64. */ + for (string::const_iterator I = S.begin(); I < S.end(); I += 3) + { + uint8_t Bits[3] = {0,0,0}; + Bits[0] = I[0]; + if (I + 1 < S.end()) + Bits[1] = I[1]; + if (I + 2 < S.end()) + Bits[2] = I[2]; + + Final += tbl[Bits[0] >> 2]; + Final += tbl[((Bits[0] & 3) << 4) + (Bits[1] >> 4)]; + + if (I + 1 >= S.end()) + break; + + Final += tbl[((Bits[1] & 0xf) << 2) + (Bits[2] >> 6)]; + + if (I + 2 >= S.end()) + break; + + Final += tbl[Bits[2] & 0x3f]; + } + + /* Apply the padding elements, this tells how many bytes the remote + end should discard */ + if (S.length() % 3 == 2) + Final += '='; + if (S.length() % 3 == 1) + Final += "=="; + + return Final; +} + /*}}}*/ +// stringcmp - Arbitrary string compare /*{{{*/ +// --------------------------------------------------------------------- +/* This safely compares two non-null terminated strings of arbitrary + length */ +int stringcmp(const char *A,const char *AEnd,const char *B,const char *BEnd) +{ + for (; A != AEnd && B != BEnd; A++, B++) + if (*A != *B) + break; + + if (A == AEnd && B == BEnd) + return 0; + if (A == AEnd) + return 1; + if (B == BEnd) + return -1; + if (*A < *B) + return -1; + return 1; +} + +#if __GNUC__ >= 3 +int stringcmp(string::const_iterator A,string::const_iterator AEnd, + const char *B,const char *BEnd) +{ + for (; A != AEnd && B != BEnd; A++, B++) + if (*A != *B) + break; + + if (A == AEnd && B == BEnd) + return 0; + if (A == AEnd) + return 1; + if (B == BEnd) + return -1; + if (*A < *B) + return -1; + return 1; +} +int stringcmp(string::const_iterator A,string::const_iterator AEnd, + string::const_iterator B,string::const_iterator BEnd) +{ + for (; A != AEnd && B != BEnd; A++, B++) + if (*A != *B) + break; + + if (A == AEnd && B == BEnd) + return 0; + if (A == AEnd) + return 1; + if (B == BEnd) + return -1; + if (*A < *B) + return -1; + return 1; +} +#endif + /*}}}*/ +// stringcasecmp - Arbitrary case insensitive string compare /*{{{*/ +// --------------------------------------------------------------------- +/* */ +int stringcasecmp(const char *A,const char *AEnd,const char *B,const char *BEnd) +{ + for (; A != AEnd && B != BEnd; A++, B++) + if (tolower_ascii(*A) != tolower_ascii(*B)) + break; + + if (A == AEnd && B == BEnd) + return 0; + if (A == AEnd) + return 1; + if (B == BEnd) + return -1; + if (tolower_ascii(*A) < tolower_ascii(*B)) + return -1; + return 1; +} +#if __GNUC__ >= 3 +int stringcasecmp(string::const_iterator A,string::const_iterator AEnd, + const char *B,const char *BEnd) +{ + for (; A != AEnd && B != BEnd; A++, B++) + if (tolower_ascii(*A) != tolower_ascii(*B)) + break; + + if (A == AEnd && B == BEnd) + return 0; + if (A == AEnd) + return 1; + if (B == BEnd) + return -1; + if (tolower_ascii(*A) < tolower_ascii(*B)) + return -1; + return 1; +} +int stringcasecmp(string::const_iterator A,string::const_iterator AEnd, + string::const_iterator B,string::const_iterator BEnd) +{ + for (; A != AEnd && B != BEnd; A++, B++) + if (tolower_ascii(*A) != tolower_ascii(*B)) + break; + + if (A == AEnd && B == BEnd) + return 0; + if (A == AEnd) + return 1; + if (B == BEnd) + return -1; + if (tolower_ascii(*A) < tolower_ascii(*B)) + return -1; + return 1; +} +#endif + /*}}}*/ +// LookupTag - Lookup the value of a tag in a tagged string /*{{{*/ +// --------------------------------------------------------------------- +/* The format is like those used in package files and the method + communication system */ +std::string LookupTag(const std::string &Message, const char *TagC, const char *Default) +{ + std::string tag = std::string("\n") + TagC + ":"; + if (Default == nullptr) + Default = ""; + if (Message.length() < tag.length()) + return Default; + std::transform(tag.begin(), tag.end(), tag.begin(), tolower_ascii); + auto valuestart = Message.cbegin(); + // maybe the message starts directly with tag + if (Message[tag.length() - 2] == ':') + { + std::string lowstart = std::string("\n") + Message.substr(0, tag.length() - 1); + std::transform(lowstart.begin(), lowstart.end(), lowstart.begin(), tolower_ascii); + if (lowstart == tag) + valuestart = std::next(valuestart, tag.length() - 1); + } + // the tag is somewhere in the message + if (valuestart == Message.cbegin()) + { + auto const tagbegin = std::search(Message.cbegin(), Message.cend(), tag.cbegin(), tag.cend(), + [](char const a, char const b) { return tolower_ascii(a) == b; }); + if (tagbegin == Message.cend()) + return Default; + valuestart = std::next(tagbegin, tag.length()); + } + auto const is_whitespace = [](char const c) { return isspace_ascii(c) != 0 && c != '\n'; }; + auto const is_newline = [](char const c) { return c == '\n'; }; + std::string result; + valuestart = std::find_if_not(valuestart, Message.cend(), is_whitespace); + // is the first line of the value empty? + if (valuestart != Message.cend() && *valuestart == '\n') + { + valuestart = std::next(valuestart); + if (valuestart != Message.cend() && *valuestart == ' ') + valuestart = std::next(valuestart); + } + // extract the value over multiple lines removing trailing whitespace + while (valuestart < Message.cend()) + { + auto const linebreak = std::find_if(valuestart, Message.cend(), is_newline); + auto valueend = std::prev(linebreak); + // skip spaces at the end of the line + while (valueend > valuestart && is_whitespace(*valueend)) + valueend = std::prev(valueend); + // append found line to result + { + std::string tmp(valuestart, std::next(valueend)); + if (tmp != ".") + { + if (result.empty()) + result.assign(std::move(tmp)); + else + result.append(tmp); + } + } + // see if the value is multiline + if (linebreak == Message.cend()) + break; + valuestart = std::next(linebreak); + if (valuestart == Message.cend() || *valuestart != ' ') + break; + result.append("\n"); + // skip the space leading a multiline (Keep all other whitespaces in the value) + valuestart = std::next(valuestart); + } + auto const valueend = result.find_last_not_of("\n"); + if (valueend == std::string::npos) + result.clear(); + else + result.erase(valueend + 1); + return result; +} + /*}}}*/ +// StringToBool - Converts a string into a boolean /*{{{*/ +// --------------------------------------------------------------------- +/* This inspects the string to see if it is true or if it is false and + then returns the result. Several variants on true/false are checked. */ +int StringToBool(const string &Text,int Default) +{ + char *ParseEnd; + int Res = strtol(Text.c_str(),&ParseEnd,0); + // ensure that the entire string was converted by strtol to avoid + // failures on "apt-cache show -a 0ad" where the "0" is converted + const char *TextEnd = Text.c_str()+Text.size(); + if (ParseEnd == TextEnd && Res >= 0 && Res <= 1) + return Res; + + // Check for positives + if (strcasecmp(Text.c_str(),"no") == 0 || + strcasecmp(Text.c_str(),"false") == 0 || + strcasecmp(Text.c_str(),"without") == 0 || + strcasecmp(Text.c_str(),"off") == 0 || + strcasecmp(Text.c_str(),"disable") == 0) + return 0; + + // Check for negatives + if (strcasecmp(Text.c_str(),"yes") == 0 || + strcasecmp(Text.c_str(),"true") == 0 || + strcasecmp(Text.c_str(),"with") == 0 || + strcasecmp(Text.c_str(),"on") == 0 || + strcasecmp(Text.c_str(),"enable") == 0) + return 1; + + return Default; +} + /*}}}*/ +// TimeRFC1123 - Convert a time_t into RFC1123 format /*{{{*/ +// --------------------------------------------------------------------- +/* This converts a time_t into a string time representation that is + year 2000 compliant and timezone neutral */ +string TimeRFC1123(time_t Date, bool const NumericTimezone) +{ + struct tm Conv; + if (gmtime_r(&Date, &Conv) == NULL) + return ""; + + auto const posix = std::locale::classic(); + std::ostringstream datestr; + datestr.imbue(posix); + APT::StringView const fmt("%a, %d %b %Y %H:%M:%S"); + std::use_facet<std::time_put<char>>(posix).put( + std::ostreambuf_iterator<char>(datestr), + datestr, ' ', &Conv, fmt.data(), fmt.data() + fmt.size()); + if (NumericTimezone) + datestr << " +0000"; + else + datestr << " GMT"; + return datestr.str(); +} + /*}}}*/ +// ReadMessages - Read messages from the FD /*{{{*/ +// --------------------------------------------------------------------- +/* This pulls full messages from the input FD into the message buffer. + It assumes that messages will not pause during transit so no + fancy buffering is used. + + In particular: this reads blocks from the input until it believes + that it's run out of input text. Each block is terminated by a + double newline ('\n' followed by '\n'). + */ +bool ReadMessages(int Fd, vector<string> &List) +{ + char Buffer[64000]; + // Represents any left-over from the previous iteration of the + // parse loop. (i.e., if a message is split across the end + // of the buffer, it goes here) + string PartialMessage; + + do { + int const Res = read(Fd, Buffer, sizeof(Buffer)); + if (Res < 0 && errno == EINTR) + continue; + + // process we read from has died + if (Res == 0) + return false; + + // No data +#if EAGAIN != EWOULDBLOCK + if (Res < 0 && (errno == EAGAIN || errno == EWOULDBLOCK)) +#else + if (Res < 0 && errno == EAGAIN) +#endif + return true; + if (Res < 0) + return false; + + // extract the message(s) from the buffer + char const *Start = Buffer; + char const * const End = Buffer + Res; + + char const * NL = (char const *) memchr(Start, '\n', End - Start); + if (NL == NULL) + { + // end of buffer: store what we have so far and read new data in + PartialMessage.append(Start, End - Start); + Start = End; + } + else + ++NL; + + if (PartialMessage.empty() == false && Start < End) + { + // if we start with a new line, see if the partial message we have ended with one + // so that we properly detect records ending between two read() runs + // cases are: \n|\n , \r\n|\r\n and \r\n\r|\n + // the case \r|\n\r\n is handled by the usual double-newline handling + if ((NL - Start) == 1 || ((NL - Start) == 2 && *Start == '\r')) + { + if (APT::String::Endswith(PartialMessage, "\n") || APT::String::Endswith(PartialMessage, "\r\n\r")) + { + PartialMessage.erase(PartialMessage.find_last_not_of("\r\n") + 1); + List.push_back(PartialMessage); + PartialMessage.clear(); + while (NL < End && (*NL == '\n' || *NL == '\r')) ++NL; + Start = NL; + } + } + } + + while (Start < End) { + char const * NL2 = (char const *) memchr(NL, '\n', End - NL); + if (NL2 == NULL) + { + // end of buffer: store what we have so far and read new data in + PartialMessage.append(Start, End - Start); + break; + } + ++NL2; + + // did we find a double newline? + if ((NL2 - NL) == 1 || ((NL2 - NL) == 2 && *NL == '\r')) + { + PartialMessage.append(Start, NL2 - Start); + PartialMessage.erase(PartialMessage.find_last_not_of("\r\n") + 1); + List.push_back(PartialMessage); + PartialMessage.clear(); + while (NL2 < End && (*NL2 == '\n' || *NL2 == '\r')) ++NL2; + Start = NL2; + } + NL = NL2; + } + + // we have read at least one complete message and nothing left + if (PartialMessage.empty() == true) + return true; + + if (WaitFd(Fd) == false) + return false; + } while (true); +} + /*}}}*/ +// MonthConv - Converts a month string into a number /*{{{*/ +// --------------------------------------------------------------------- +/* This was lifted from the boa webserver which lifted it from 'wn-v1.07' + Made it a bit more robust with a few tolower_ascii though. */ +static int MonthConv(char const * const Month) +{ + switch (tolower_ascii(*Month)) + { + case 'a': + return tolower_ascii(Month[1]) == 'p'?3:7; + case 'd': + return 11; + case 'f': + return 1; + case 'j': + if (tolower_ascii(Month[1]) == 'a') + return 0; + return tolower_ascii(Month[2]) == 'n'?5:6; + case 'm': + return tolower_ascii(Month[2]) == 'r'?2:4; + case 'n': + return 10; + case 'o': + return 9; + case 's': + return 8; + + // Pretend it is January.. + default: + return 0; + } +} + /*}}}*/ +// timegm - Internal timegm if the gnu version is not available /*{{{*/ +// --------------------------------------------------------------------- +/* Converts struct tm to time_t, assuming the data in tm is UTC rather + than local timezone (mktime assumes the latter). + + This function is a nonstandard GNU extension that is also present on + the BSDs and maybe other systems. For others we follow the advice of + the manpage of timegm and use his portable replacement. */ +#ifndef HAVE_TIMEGM +static time_t timegm(struct tm *t) +{ + char *tz = getenv("TZ"); + setenv("TZ", "", 1); + tzset(); + time_t ret = mktime(t); + if (tz) + setenv("TZ", tz, 1); + else + unsetenv("TZ"); + tzset(); + return ret; +} +#endif + /*}}}*/ +// RFC1123StrToTime - Converts an HTTP1.1 full date strings into a time_t /*{{{*/ +// --------------------------------------------------------------------- +/* tries to parses a full date as specified in RFC7231 §7.1.1.1 + with one exception: HTTP/1.1 valid dates need to have GMT as timezone. + As we encounter dates from UTC or with a numeric timezone in other places, + we allow them here to to be able to reuse the method. Either way, a date + must be in UTC or parsing will fail. Previous implementations of this + method used to ignore the timezone and assume always UTC. */ +bool RFC1123StrToTime(std::string const &str,time_t &time) +{ + unsigned short day = 0; + signed int year = 0; // yes, Y23K problem – we going to worry then… + std::string weekday, month, datespec, timespec, zone; + std::istringstream ss(str); + auto const &posix = std::locale::classic(); + ss.imbue(posix); + ss >> weekday; + // we only superficially check weekday, mostly to avoid accepting localized + // weekdays here and take only its length to decide which datetime format we + // encounter here. The date isn't stored. + std::transform(weekday.begin(), weekday.end(), weekday.begin(), ::tolower); + std::array<char const * const, 7> c_weekdays = {{ "sun", "mon", "tue", "wed", "thu", "fri", "sat" }}; + if (std::find(c_weekdays.begin(), c_weekdays.end(), weekday.substr(0,3)) == c_weekdays.end()) + return false; + + switch (weekday.length()) + { + case 4: + // Sun, 06 Nov 1994 08:49:37 GMT ; RFC 822, updated by RFC 1123 + if (weekday[3] != ',') + return false; + ss >> day >> month >> year >> timespec >> zone; + break; + case 3: + // Sun Nov 6 08:49:37 1994 ; ANSI C's asctime() format + ss >> month >> day >> timespec >> year; + zone = "UTC"; + break; + case 0: + case 1: + case 2: + return false; + default: + // Sunday, 06-Nov-94 08:49:37 GMT ; RFC 850, obsoleted by RFC 1036 + if (weekday[weekday.length() - 1] != ',') + return false; + ss >> datespec >> timespec >> zone; + auto const expldate = VectorizeString(datespec, '-'); + if (expldate.size() != 3) + return false; + try { + size_t pos; + day = std::stoi(expldate[0], &pos); + if (pos != expldate[0].length()) + return false; + year = 1900 + std::stoi(expldate[2], &pos); + if (pos != expldate[2].length()) + return false; + strprintf(datespec, "%.4d-%.2d-%.2d", year, MonthConv(expldate[1].c_str()) + 1, day); + } catch (...) { + return false; + } + break; + } + + if (ss.fail() || ss.bad() || !ss.eof()) + return false; + + if (zone != "GMT" && zone != "UTC" && zone != "Z") // RFC 822 + { + // numeric timezones as a should of RFC 1123 and generally preferred + try { + size_t pos; + auto const z = std::stoi(zone, &pos); + if (z != 0 || pos != zone.length()) + return false; + } catch (...) { + return false; + } + } + + if (datespec.empty()) + { + if (month.empty()) + return false; + strprintf(datespec, "%.4d-%.2d-%.2d", year, MonthConv(month.c_str()) + 1, day); + } + + std::string const datetime = datespec + ' ' + timespec; + struct tm Tm; + if (strptime(datetime.c_str(), "%Y-%m-%d %H:%M:%S", &Tm) == nullptr) + return false; + time = timegm(&Tm); + return true; +} + /*}}}*/ +// FTPMDTMStrToTime - Converts a ftp modification date into a time_t /*{{{*/ +// --------------------------------------------------------------------- +/* */ +bool FTPMDTMStrToTime(const char* const str,time_t &time) +{ + struct tm Tm; + // MDTM includes no whitespaces but recommend and ignored by strptime + if (strptime(str, "%Y %m %d %H %M %S", &Tm) == NULL) + return false; + + time = timegm(&Tm); + return true; +} + /*}}}*/ +// StrToNum - Convert a fixed length string to a number /*{{{*/ +// --------------------------------------------------------------------- +/* This is used in decoding the crazy fixed length string headers in + tar and ar files. */ +bool StrToNum(const char *Str,unsigned long &Res,unsigned Len,unsigned Base) +{ + unsigned long long BigRes; + if (not StrToNum(Str, BigRes, Len, Base)) + return false; + + if (std::numeric_limits<unsigned long>::max() < BigRes) + return false; + + Res = BigRes; + return true; +} + /*}}}*/ +// StrToNum - Convert a fixed length string to a number /*{{{*/ +// --------------------------------------------------------------------- +/* This is used in decoding the crazy fixed length string headers in + tar and ar files. */ +bool StrToNum(const char *Str,unsigned long long &Res,unsigned Len,unsigned Base) +{ + char S[30]; + if (Len >= sizeof(S)) + return false; + memcpy(S,Str,Len); + S[Len] = 0; + + // All spaces is a zero + Res = 0; + unsigned I; + for (I = 0; S[I] == ' '; ++I); + if (S[I] == 0) + return true; + if (S[I] == '-') + return false; + + char *End; + errno = 0; + Res = strtoull(S,&End,Base); + return not (End == S || errno != 0); +} + /*}}}*/ + +// Base256ToNum - Convert a fixed length binary to a number /*{{{*/ +// --------------------------------------------------------------------- +/* This is used in decoding the 256bit encoded fixed length fields in + tar files */ +bool Base256ToNum(const char *Str,unsigned long long &Res,unsigned int Len) +{ + if ((Str[0] & 0x80) == 0) + return false; + else + { + Res = Str[0] & 0x7F; + for(unsigned int i = 1; i < Len; ++i) + Res = (Res<<8) + Str[i]; + return true; + } +} + /*}}}*/ +// Base256ToNum - Convert a fixed length binary to a number /*{{{*/ +// --------------------------------------------------------------------- +/* This is used in decoding the 256bit encoded fixed length fields in + tar files */ +bool Base256ToNum(const char *Str,unsigned long &Res,unsigned int Len) +{ + unsigned long long Num = 0; + bool rc; + + rc = Base256ToNum(Str, Num, Len); + // rudimentary check for overflow (Res = ulong, Num = ulonglong) + Res = Num; + if (Res != Num) + return false; + + return rc; +} + /*}}}*/ +// HexDigit - Convert a hex character into an integer /*{{{*/ +// --------------------------------------------------------------------- +/* Helper for Hex2Num */ +static int HexDigit(int c) +{ + if (c >= '0' && c <= '9') + return c - '0'; + if (c >= 'a' && c <= 'f') + return c - 'a' + 10; + if (c >= 'A' && c <= 'F') + return c - 'A' + 10; + return -1; +} + /*}}}*/ +// Hex2Num - Convert a long hex number into a buffer /*{{{*/ +// --------------------------------------------------------------------- +/* The length of the buffer must be exactly 1/2 the length of the string. */ +bool Hex2Num(const APT::StringView Str,unsigned char *Num,unsigned int Length) +{ + if (Str.length() != Length*2) + return false; + + // Convert each digit. We store it in the same order as the string + int J = 0; + for (auto I = Str.begin(); I != Str.end();J++, I += 2) + { + int first_half = HexDigit(I[0]); + int second_half; + if (first_half < 0) + return false; + + second_half = HexDigit(I[1]); + if (second_half < 0) + return false; + Num[J] = first_half << 4; + Num[J] += second_half; + } + + return true; +} + /*}}}*/ +// TokSplitString - Split a string up by a given token /*{{{*/ +// --------------------------------------------------------------------- +/* This is intended to be a faster splitter, it does not use dynamic + memories. Input is changed to insert nulls at each token location. */ +bool TokSplitString(char Tok,char *Input,char **List, + unsigned long ListMax) +{ + // Strip any leading spaces + char *Start = Input; + char *Stop = Start + strlen(Start); + for (; *Start != 0 && isspace(*Start) != 0; Start++); + + unsigned long Count = 0; + char *Pos = Start; + while (Pos != Stop) + { + // Skip to the next Token + for (; Pos != Stop && *Pos != Tok; Pos++); + + // Back remove spaces + char *End = Pos; + for (; End > Start && (End[-1] == Tok || isspace(End[-1]) != 0); End--); + *End = 0; + + List[Count++] = Start; + if (Count >= ListMax) + { + List[Count-1] = 0; + return false; + } + + // Advance pos + for (; Pos != Stop && (*Pos == Tok || isspace(*Pos) != 0 || *Pos == 0); Pos++); + Start = Pos; + } + + List[Count] = 0; + return true; +} + /*}}}*/ +// VectorizeString - Split a string up into a vector of strings /*{{{*/ +// --------------------------------------------------------------------- +/* This can be used to split a given string up into a vector, so the + propose is the same as in the method above and this one is a bit slower + also, but the advantage is that we have an iteratable vector */ +vector<string> VectorizeString(string const &haystack, char const &split) +{ + vector<string> exploded; + if (haystack.empty() == true) + return exploded; + string::const_iterator start = haystack.begin(); + string::const_iterator end = start; + do { + for (; end != haystack.end() && *end != split; ++end); + exploded.push_back(string(start, end)); + start = end + 1; + } while (end != haystack.end() && (++end) != haystack.end()); + return exploded; +} + /*}}}*/ +// StringSplit - split a string into a string vector by token /*{{{*/ +// --------------------------------------------------------------------- +/* See header for details. + */ +vector<string> StringSplit(std::string const &s, std::string const &sep, + unsigned int maxsplit) +{ + vector<string> split; + size_t start, pos; + + // no separator given, this is bogus + if(sep.size() == 0) + return split; + + start = pos = 0; + while (pos != string::npos) + { + pos = s.find(sep, start); + split.push_back(s.substr(start, pos-start)); + + // if maxsplit is reached, the remaining string is the last item + if(split.size() >= maxsplit) + { + split[split.size()-1] = s.substr(start); + break; + } + start = pos+sep.size(); + } + return split; +} + /*}}}*/ +// RegexChoice - Simple regex list/list matcher /*{{{*/ +// --------------------------------------------------------------------- +/* */ +unsigned long RegexChoice(RxChoiceList *Rxs,const char **ListBegin, + const char **ListEnd) +{ + for (RxChoiceList *R = Rxs; R->Str != 0; R++) + R->Hit = false; + + unsigned long Hits = 0; + for (; ListBegin < ListEnd; ++ListBegin) + { + // Check if the name is a regex + const char *I; + bool Regex = true; + for (I = *ListBegin; *I != 0; I++) + if (*I == '.' || *I == '?' || *I == '*' || *I == '|') + break; + if (*I == 0) + Regex = false; + + // Compile the regex pattern + regex_t Pattern; + if (Regex == true) + if (regcomp(&Pattern,*ListBegin,REG_EXTENDED | REG_ICASE | + REG_NOSUB) != 0) + Regex = false; + + // Search the list + bool Done = false; + for (RxChoiceList *R = Rxs; R->Str != 0; R++) + { + if (R->Str[0] == 0) + continue; + + if (strcasecmp(R->Str,*ListBegin) != 0) + { + if (Regex == false) + continue; + if (regexec(&Pattern,R->Str,0,0,0) != 0) + continue; + } + Done = true; + + if (R->Hit == false) + Hits++; + + R->Hit = true; + } + + if (Regex == true) + regfree(&Pattern); + + if (Done == false) + _error->Warning(_("Selection %s not found"),*ListBegin); + } + + return Hits; +} + /*}}}*/ +// {str,io}printf - C format string outputter to C++ strings/iostreams /*{{{*/ +// --------------------------------------------------------------------- +/* This is used to make the internationalization strings easier to translate + and to allow reordering of parameters */ +bool iovprintf(std::ostream &out, const char *format, + va_list &args, ssize_t &size) { + auto S = std::unique_ptr<char,decltype(&free)>{static_cast<char*>(malloc(size)), &free}; + ssize_t const n = vsnprintf(S.get(), size, format, args); + if (n > -1 && n < size) { + out << S.get(); + return true; + } else { + if (n > -1) + size = n + 1; + else + size *= 2; + } + return false; +} +void ioprintf(ostream &out,const char *format,...) +{ + va_list args; + ssize_t size = 400; + while (true) { + bool ret; + va_start(args,format); + ret = iovprintf(out, format, args, size); + va_end(args); + if (ret == true) + return; + } +} +void strprintf(string &out,const char *format,...) +{ + va_list args; + ssize_t size = 400; + std::ostringstream outstr; + while (true) { + bool ret; + va_start(args,format); + ret = iovprintf(outstr, format, args, size); + va_end(args); + if (ret == true) + break; + } + out = outstr.str(); +} + /*}}}*/ +// safe_snprintf - Safer snprintf /*{{{*/ +// --------------------------------------------------------------------- +/* This is a snprintf that will never (ever) go past 'End' and returns a + pointer to the end of the new string. The returned string is always null + terminated unless Buffer == end. This is a better alterantive to using + consecutive snprintfs. */ +char *safe_snprintf(char *Buffer,char *End,const char *Format,...) +{ + va_list args; + int Did; + + if (End <= Buffer) + return End; + va_start(args,Format); + Did = vsnprintf(Buffer,End - Buffer,Format,args); + va_end(args); + + if (Did < 0 || Buffer + Did > End) + return End; + return Buffer + Did; +} + /*}}}*/ +// StripEpoch - Remove the version "epoch" from a version string /*{{{*/ +// --------------------------------------------------------------------- +string StripEpoch(const string &VerStr) +{ + size_t i = VerStr.find(":"); + if (i == string::npos) + return VerStr; + return VerStr.substr(i+1); +} + /*}}}*/ + +// tolower_ascii - tolower() function that ignores the locale /*{{{*/ +// --------------------------------------------------------------------- +/* This little function is the most called method we have and tries + therefore to do the absolute minimum - and is notable faster than + standard tolower/toupper and as a bonus avoids problems with different + locales - we only operate on ascii chars anyway. */ +#undef tolower_ascii +int tolower_ascii(int const c) APT_PURE APT_COLD; +int tolower_ascii(int const c) +{ + return tolower_ascii_inline(c); +} + /*}}}*/ + +// isspace_ascii - isspace() function that ignores the locale /*{{{*/ +// --------------------------------------------------------------------- +/* This little function is one of the most called methods we have and tries + therefore to do the absolute minimum - and is notable faster than + standard isspace() and as a bonus avoids problems with different + locales - we only operate on ascii chars anyway. */ +#undef isspace_ascii +int isspace_ascii(int const c) APT_PURE APT_COLD; +int isspace_ascii(int const c) +{ + return isspace_ascii_inline(c); +} + /*}}}*/ + +// CheckDomainList - See if Host is in a , separate list /*{{{*/ +// --------------------------------------------------------------------- +/* The domain list is a comma separate list of domains that are suffix + matched against the argument */ +bool CheckDomainList(const string &Host,const string &List) +{ + string::const_iterator Start = List.begin(); + for (string::const_iterator Cur = List.begin(); Cur <= List.end(); ++Cur) + { + if (Cur < List.end() && *Cur != ',') + continue; + + // Match the end of the string.. + if ((Host.size() >= (unsigned)(Cur - Start)) && + Cur - Start != 0 && + stringcasecmp(Host.end() - (Cur - Start),Host.end(),Start,Cur) == 0) + return true; + + Start = Cur + 1; + } + return false; +} + /*}}}*/ +// strv_length - Return the length of a NULL-terminated string array /*{{{*/ +// --------------------------------------------------------------------- +/* */ +size_t strv_length(const char **str_array) +{ + size_t i; + for (i=0; str_array[i] != NULL; i++) + /* nothing */ + ; + return i; +} + /*}}}*/ +// DeEscapeString - unescape (\0XX and \xXX) from a string /*{{{*/ +// --------------------------------------------------------------------- +/* */ +string DeEscapeString(const string &input) +{ + char tmp[3]; + string::const_iterator it; + string output; + for (it = input.begin(); it != input.end(); ++it) + { + // just copy non-escape chars + if (*it != '\\') + { + output += *it; + continue; + } + + // deal with double escape + if (*it == '\\' && + (it + 1 < input.end()) && it[1] == '\\') + { + // copy + output += *it; + // advance iterator one step further + ++it; + continue; + } + + // ensure we have a char to read + if (it + 1 == input.end()) + continue; + + // read it + ++it; + switch (*it) + { + case '0': + if (it + 2 < input.end()) { + tmp[0] = it[1]; + tmp[1] = it[2]; + tmp[2] = 0; + output += (char)strtol(tmp, 0, 8); + it += 2; + } else { + // FIXME: raise exception here? + } + break; + case 'x': + if (it + 2 < input.end()) { + tmp[0] = it[1]; + tmp[1] = it[2]; + tmp[2] = 0; + output += (char)strtol(tmp, 0, 16); + it += 2; + } else { + // FIXME: raise exception here? + } + break; + default: + // FIXME: raise exception here? + break; + } + } + return output; +} + /*}}}*/ +// URI::CopyFrom - Copy from an object /*{{{*/ +// --------------------------------------------------------------------- +/* This parses the URI into all of its components */ +void URI::CopyFrom(const string &U) +{ + string::const_iterator I = U.begin(); + + // Locate the first colon, this separates the scheme + for (; I < U.end() && *I != ':' ; ++I); + string::const_iterator FirstColon = I; + + /* Determine if this is a host type URI with a leading double // + and then search for the first single / */ + string::const_iterator SingleSlash = I; + if (I + 3 < U.end() && I[1] == '/' && I[2] == '/') + SingleSlash += 3; + + /* Find the / indicating the end of the hostname, ignoring /'s in the + square brackets */ + bool InBracket = false; + for (; SingleSlash < U.end() && (*SingleSlash != '/' || InBracket == true); ++SingleSlash) + { + if (*SingleSlash == '[') + InBracket = true; + if (InBracket == true && *SingleSlash == ']') + InBracket = false; + } + + if (SingleSlash > U.end()) + SingleSlash = U.end(); + + // We can now write the access and path specifiers + Access.assign(U.begin(),FirstColon); + if (SingleSlash != U.end()) + Path.assign(SingleSlash,U.end()); + if (Path.empty() == true) + Path = "/"; + + // Now we attempt to locate a user:pass@host fragment + if (FirstColon + 2 <= U.end() && FirstColon[1] == '/' && FirstColon[2] == '/') + FirstColon += 3; + else + FirstColon += 1; + if (FirstColon >= U.end()) + return; + + if (FirstColon > SingleSlash) + FirstColon = SingleSlash; + + // Find the colon... + I = FirstColon + 1; + if (I > SingleSlash) + I = SingleSlash; + + // Search for the @ separating user:pass from host + auto const RevAt = std::find( + std::string::const_reverse_iterator(SingleSlash), + std::string::const_reverse_iterator(I), '@'); + string::const_iterator const At = RevAt.base() == I ? SingleSlash : std::prev(RevAt.base()); + // and then look for the colon between user and pass + string::const_iterator const SecondColon = std::find(I, At, ':'); + + // Now write the host and user/pass + if (At == SingleSlash) + { + if (FirstColon < SingleSlash) + Host.assign(FirstColon,SingleSlash); + } + else + { + Host.assign(At+1,SingleSlash); + // username and password must be encoded (RFC 3986) + User.assign(DeQuoteString(FirstColon,SecondColon)); + if (SecondColon < At) + Password.assign(DeQuoteString(SecondColon+1,At)); + } + + // Now we parse the RFC 2732 [] hostnames. + unsigned long PortEnd = 0; + InBracket = false; + for (unsigned I = 0; I != Host.length();) + { + if (Host[I] == '[') + { + InBracket = true; + Host.erase(I,1); + continue; + } + + if (InBracket == true && Host[I] == ']') + { + InBracket = false; + Host.erase(I,1); + PortEnd = I; + continue; + } + I++; + } + + // Tsk, weird. + if (InBracket == true) + { + Host.clear(); + return; + } + + // Now we parse off a port number from the hostname + Port = 0; + string::size_type Pos = Host.rfind(':'); + if (Pos == string::npos || Pos < PortEnd) + return; + + Port = atoi(string(Host,Pos+1).c_str()); + Host.assign(Host,0,Pos); +} + /*}}}*/ +// URI::operator string - Convert the URI to a string /*{{{*/ +// --------------------------------------------------------------------- +/* */ +URI::operator string() +{ + std::stringstream Res; + + if (Access.empty() == false) + Res << Access << ':'; + + if (Host.empty() == false) + { + if (Access.empty() == false) + Res << "//"; + + if (User.empty() == false) + { + // FIXME: Technically userinfo is permitted even less + // characters than these, but this is not conveniently + // expressed with a denylist. + Res << QuoteString(User, ":/?#[]@"); + if (Password.empty() == false) + Res << ":" << QuoteString(Password, ":/?#[]@"); + Res << "@"; + } + + // Add RFC 2732 escaping characters + if (Access.empty() == false && Host.find_first_of("/:") != string::npos) + Res << '[' << Host << ']'; + else + Res << Host; + + if (Port != 0) + Res << ':' << std::to_string(Port); + } + + if (Path.empty() == false) + { + if (Path[0] != '/') + Res << "/" << Path; + else + Res << Path; + } + + return Res.str(); +} + /*}}}*/ +// URI::SiteOnly - Return the schema and site for the URI /*{{{*/ +string URI::SiteOnly(const string &URI) +{ + ::URI U(URI); + U.User.clear(); + U.Password.clear(); + U.Path.clear(); + return U; +} + /*}}}*/ +// URI::ArchiveOnly - Return the schema, site and cleaned path for the URI /*{{{*/ +string URI::ArchiveOnly(const string &URI) +{ + ::URI U(URI); + U.User.clear(); + U.Password.clear(); + if (U.Path.empty() == false && U.Path[U.Path.length() - 1] == '/') + U.Path.erase(U.Path.length() - 1); + return U; +} + /*}}}*/ +// URI::NoUserPassword - Return the schema, site and path for the URI /*{{{*/ +string URI::NoUserPassword(const string &URI) +{ + ::URI U(URI); + U.User.clear(); + U.Password.clear(); + return U; +} + /*}}}*/ diff --git a/apt-pkg/contrib/strutl.h b/apt-pkg/contrib/strutl.h new file mode 100644 index 0000000..1fdc8dc --- /dev/null +++ b/apt-pkg/contrib/strutl.h @@ -0,0 +1,252 @@ +// -*- mode: cpp; mode: fold -*- +// SPDX-License-Identifier: GPL-2.0+ +// Description /*{{{*/ +/* ###################################################################### + + String Util - These are some useful string functions + + _strstrip is a function to remove whitespace from the front and end + of a string. + + This file had this historic note, but now includes further changes + under the GPL-2.0+: + + This source is placed in the Public Domain, do with it what you will + It was originally written by Jason Gunthorpe <jgg@gpu.srv.ualberta.ca> + + ##################################################################### */ + /*}}}*/ +#ifndef STRUTL_H +#define STRUTL_H + +#include <cstring> +#include <iostream> +#include <limits> +#include <string> +#include <vector> +#include <apt-pkg/string_view.h> +#include <stddef.h> +#include <time.h> + +#include "macros.h" + + +namespace APT { + namespace String { + APT_PUBLIC std::string Strip(const std::string &s); + APT_PUBLIC bool Endswith(const std::string &s, const std::string &ending); + APT_PUBLIC bool Startswith(const std::string &s, const std::string &starting); + APT_PUBLIC std::string Join(std::vector<std::string> list, const std::string &sep); + // Returns string display length honoring multi-byte characters + APT_PUBLIC size_t DisplayLength(StringView str); + } +} + + +APT_PUBLIC bool UTF8ToCodeset(const char *codeset, const std::string &orig, std::string *dest); +APT_PUBLIC char *_strstrip(char *String); +APT_PUBLIC char *_strrstrip(char *String); // right strip only +APT_DEPRECATED_MSG("Use SubstVar to avoid memory headaches") APT_PUBLIC char *_strtabexpand(char *String,size_t Len); +APT_PUBLIC bool ParseQuoteWord(const char *&String,std::string &Res); +APT_PUBLIC bool ParseCWord(const char *&String,std::string &Res); +APT_PUBLIC std::string QuoteString(const std::string &Str,const char *Bad); +APT_PUBLIC std::string DeQuoteString(const std::string &Str); +APT_PUBLIC std::string DeQuoteString(std::string::const_iterator const &begin, std::string::const_iterator const &end); + +// unescape (\0XX and \xXX) from a string +APT_PUBLIC std::string DeEscapeString(const std::string &input); + +APT_PUBLIC std::string SizeToStr(double Bytes); +APT_PUBLIC std::string TimeToStr(unsigned long Sec); +APT_PUBLIC std::string Base64Encode(const std::string &Str); +APT_PUBLIC std::string OutputInDepth(const unsigned long Depth, const char* Separator=" "); +APT_PUBLIC std::string URItoFileName(const std::string &URI); +/** returns a datetime string as needed by HTTP/1.1 and Debian files. + * + * Note: The date will always be represented in a UTC timezone + * + * @param Date to be represented as a string + * @param NumericTimezone is preferred in general, but HTTP/1.1 requires the use + * of GMT as timezone instead. \b true means that the timezone should be denoted + * as "+0000" while \b false uses "GMT". + */ +APT_PUBLIC std::string TimeRFC1123(time_t Date, bool const NumericTimezone); +/** parses time as needed by HTTP/1.1 and Debian files. + * + * HTTP/1.1 prefers dates in RFC1123 format (but the other two obsolete date formats + * are supported to) and e.g. Release files use the same format in Date & Valid-Until + * fields. + * + * Note: datetime strings need to be in UTC timezones (GMT, UTC, Z, +/-0000) to be + * parsed. Other timezones will be rejected as invalid. Previous implementations + * accepted other timezones, but treated them as UTC. + * + * @param str is the datetime string to parse + * @param[out] time will be the seconds since epoch of the given datetime if + * parsing is successful, undefined otherwise. + * @return \b true if parsing was successful, otherwise \b false. + */ +APT_PUBLIC bool RFC1123StrToTime(const std::string &str,time_t &time) APT_MUSTCHECK; +APT_PUBLIC bool FTPMDTMStrToTime(const char* const str,time_t &time) APT_MUSTCHECK; +APT_PUBLIC std::string LookupTag(const std::string &Message,const char *Tag,const char *Default = 0); +APT_PUBLIC int StringToBool(const std::string &Text,int Default = -1); +APT_PUBLIC bool ReadMessages(int Fd, std::vector<std::string> &List); +APT_PUBLIC bool StrToNum(const char *Str,unsigned long &Res,unsigned Len,unsigned Base = 0); +APT_PUBLIC bool StrToNum(const char *Str,unsigned long long &Res,unsigned Len,unsigned Base = 0); +APT_PUBLIC bool Base256ToNum(const char *Str,unsigned long &Res,unsigned int Len); +APT_PUBLIC bool Base256ToNum(const char *Str,unsigned long long &Res,unsigned int Len); +APT_PUBLIC bool Hex2Num(const APT::StringView Str,unsigned char *Num,unsigned int Length); +// input changing string split +APT_PUBLIC bool TokSplitString(char Tok,char *Input,char **List, + unsigned long ListMax); + +// split a given string by a char +APT_PUBLIC std::vector<std::string> VectorizeString(std::string const &haystack, char const &split) APT_PURE; + +/* \brief Return a vector of strings from string "input" where "sep" + * is used as the delimiter string. + * + * \param input The input string. + * + * \param sep The separator to use. + * + * \param maxsplit (optional) The maximum amount of splitting that + * should be done . + * + * The optional "maxsplit" argument can be used to limit the splitting, + * if used the string is only split on maxsplit places and the last + * item in the vector contains the remainder string. + */ +APT_PUBLIC std::vector<std::string> StringSplit(std::string const &input, + std::string const &sep, + unsigned int maxsplit=std::numeric_limits<unsigned int>::max()) APT_PURE; + + +APT_HIDDEN bool iovprintf(std::ostream &out, const char *format, va_list &args, ssize_t &size); +APT_PUBLIC void ioprintf(std::ostream &out,const char *format,...) APT_PRINTF(2); +APT_PUBLIC void strprintf(std::string &out,const char *format,...) APT_PRINTF(2); +APT_PUBLIC char *safe_snprintf(char *Buffer,char *End,const char *Format,...) APT_PRINTF(3); +APT_PUBLIC bool CheckDomainList(const std::string &Host, const std::string &List); + +/* Do some compat mumbo jumbo */ +#define tolower_ascii tolower_ascii_inline +#define isspace_ascii isspace_ascii_inline + +APT_PURE APT_HOT +static inline int tolower_ascii_unsafe(int const c) +{ + return c | 0x20; +} +APT_PURE APT_HOT +static inline int tolower_ascii_inline(int const c) +{ + return (c >= 'A' && c <= 'Z') ? c + 32 : c; +} +APT_PURE APT_HOT +static inline int isspace_ascii_inline(int const c) +{ + // 9='\t',10='\n',11='\v',12='\f',13='\r',32=' ' + return (c >= 9 && c <= 13) || c == ' '; +} +APT_PURE APT_HOT +static inline int islower_ascii(int const c) +{ + return c >= 'a' && c <= 'z'; +} +APT_PURE APT_HOT +static inline int isupper_ascii(int const c) +{ + return c >= 'A' && c <= 'Z'; +} +APT_PURE APT_HOT +static inline int isalpha_ascii(int const c) +{ + return isupper_ascii(c) || islower_ascii(c); +} + +APT_PUBLIC std::string StripEpoch(const std::string &VerStr); + +#define APT_MKSTRCMP(name,func) \ +inline APT_PURE int name(const char *A,const char *B) {return func(A,A+strlen(A),B,B+strlen(B));} \ +inline APT_PURE int name(const char *A,const char *AEnd,const char *B) {return func(A,AEnd,B,B+strlen(B));} \ +inline APT_PURE int name(const std::string& A,const char *B) {return func(A.c_str(),A.c_str()+A.length(),B,B+strlen(B));} \ +inline APT_PURE int name(const std::string& A,const std::string& B) {return func(A.c_str(),A.c_str()+A.length(),B.c_str(),B.c_str()+B.length());} \ +inline APT_PURE int name(const std::string& A,const char *B,const char *BEnd) {return func(A.c_str(),A.c_str()+A.length(),B,BEnd);} + +#define APT_MKSTRCMP2(name,func) \ +inline APT_PURE int name(const char *A,const char *AEnd,const char *B) {return func(A,AEnd,B,B+strlen(B));} \ +inline APT_PURE int name(const std::string& A,const char *B) {return func(A.begin(),A.end(),B,B+strlen(B));} \ +inline APT_PURE int name(const std::string& A,const std::string& B) {return func(A.begin(),A.end(),B.begin(),B.end());} \ +inline APT_PURE int name(const std::string& A,const char *B,const char *BEnd) {return func(A.begin(),A.end(),B,BEnd);} + +APT_PUBLIC int APT_PURE stringcmp(const char *A,const char *AEnd,const char *B,const char *BEnd); +APT_PUBLIC int APT_PURE stringcasecmp(const char *A,const char *AEnd,const char *B,const char *BEnd); + +/* We assume that GCC 3 indicates that libstdc++3 is in use too. In that + case the definition of string::const_iterator is not the same as + const char * and we need these extra functions */ +#if __GNUC__ >= 3 +APT_PUBLIC int APT_PURE stringcmp(std::string::const_iterator A,std::string::const_iterator AEnd, + const char *B,const char *BEnd); +APT_PUBLIC int APT_PURE stringcmp(std::string::const_iterator A,std::string::const_iterator AEnd, + std::string::const_iterator B,std::string::const_iterator BEnd); +APT_PUBLIC int APT_PURE stringcasecmp(std::string::const_iterator A,std::string::const_iterator AEnd, + const char *B,const char *BEnd); +APT_PUBLIC int APT_PURE stringcasecmp(std::string::const_iterator A,std::string::const_iterator AEnd, + std::string::const_iterator B,std::string::const_iterator BEnd); + +inline APT_PURE int stringcmp(std::string::const_iterator A,std::string::const_iterator Aend,const char *B) {return stringcmp(A,Aend,B,B+strlen(B));} +inline APT_PURE int stringcasecmp(std::string::const_iterator A,std::string::const_iterator Aend,const char *B) {return stringcasecmp(A,Aend,B,B+strlen(B));} +#endif + +APT_MKSTRCMP2(stringcmp,stringcmp) +APT_MKSTRCMP2(stringcasecmp,stringcasecmp) + +// Return the length of a NULL-terminated string array +APT_PUBLIC size_t APT_PURE strv_length(const char **str_array); + + +inline const char *DeNull(const char *s) {return (s == 0?"(null)":s);} + +class APT_PUBLIC URI +{ + void CopyFrom(const std::string &From); + + public: + + std::string Access; + std::string User; + std::string Password; + std::string Host; + std::string Path; + unsigned int Port; + + operator std::string(); + inline void operator =(const std::string &From) {CopyFrom(From);} + inline bool empty() {return Access.empty();}; + static std::string SiteOnly(const std::string &URI); + static std::string ArchiveOnly(const std::string &URI); + static std::string NoUserPassword(const std::string &URI); + + explicit URI(std::string Path) { CopyFrom(Path); } + URI() : Port(0) {} +}; + +struct SubstVar +{ + const char *Subst; + const std::string *Contents; +}; +APT_PUBLIC std::string SubstVar(std::string Str,const struct SubstVar *Vars); +APT_PUBLIC std::string SubstVar(const std::string &Str,const std::string &Subst,const std::string &Contents); + +struct RxChoiceList +{ + void *UserData; + const char *Str; + bool Hit; +}; +APT_PUBLIC unsigned long RegexChoice(RxChoiceList *Rxs,const char **ListBegin, + const char **ListEnd); + +#endif diff --git a/apt-pkg/contrib/weakptr.h b/apt-pkg/contrib/weakptr.h new file mode 100644 index 0000000..8de727d --- /dev/null +++ b/apt-pkg/contrib/weakptr.h @@ -0,0 +1,64 @@ +/* weakptr.h - An object which supports weak pointers. + * + * Copyright (C) 2010 Julian Andres Klode <jak@debian.org> + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, + * MA 02110-1301, USA. + */ + +#ifndef WEAK_POINTER_H +#define WEAK_POINTER_H + +#include <set> +#include <stddef.h> + +/** + * Class for objects providing support for weak pointers. + * + * This class allows for the registration of certain pointers as weak, + * which will cause them to be set to NULL when the destructor of the + * object is called. + */ +class WeakPointable { +private: + std::set<WeakPointable**> pointers; + +public: + + /** + * Add a new weak pointer. + */ + inline void AddWeakPointer(WeakPointable** weakptr) { + pointers.insert(weakptr); + } + + /** + * Remove the weak pointer from the list of weak pointers. + */ + inline void RemoveWeakPointer(WeakPointable **weakptr) { + pointers.erase(weakptr); + } + + /** + * Deconstruct the object, set all weak pointers to NULL. + */ + ~WeakPointable() { + std::set<WeakPointable**>::iterator iter = pointers.begin(); + while (iter != pointers.end()) + **(iter++) = NULL; + } +}; + +#endif // WEAK_POINTER_H |