diff options
Diffstat (limited to 'intl/icu/source/tools/toolutil/pkg_genc.cpp')
| -rw-r--r-- | intl/icu/source/tools/toolutil/pkg_genc.cpp | 1396 |
1 files changed, 1396 insertions, 0 deletions
diff --git a/intl/icu/source/tools/toolutil/pkg_genc.cpp b/intl/icu/source/tools/toolutil/pkg_genc.cpp new file mode 100644 index 0000000000..741a8a5228 --- /dev/null +++ b/intl/icu/source/tools/toolutil/pkg_genc.cpp @@ -0,0 +1,1396 @@ +// © 2016 and later: Unicode, Inc. and others. +// License & terms of use: http://www.unicode.org/copyright.html +/****************************************************************************** + * Copyright (C) 2009-2016, International Business Machines + * Corporation and others. All Rights Reserved. + ******************************************************************************* + */ +#include "unicode/utypes.h" + +#if U_PLATFORM_HAS_WIN32_API +# define VC_EXTRALEAN +# define WIN32_LEAN_AND_MEAN +# define NOUSER +# define NOSERVICE +# define NOIME +# define NOMCX +#include <windows.h> +#include <time.h> +# ifdef __GNUC__ +# define WINDOWS_WITH_GNUC +# endif +#endif + +#if U_PLATFORM_IS_LINUX_BASED && U_HAVE_ELF_H +# define U_ELF +#endif + +#ifdef U_ELF +# include <elf.h> +# if defined(ELFCLASS64) +# define U_ELF64 +# endif + /* Old elf.h headers may not have EM_X86_64, or have EM_X8664 instead. */ +# ifndef EM_X86_64 +# define EM_X86_64 62 +# endif +# define ICU_ENTRY_OFFSET 0 +#endif + +#include <stdio.h> +#include <stdlib.h> +#include "unicode/putil.h" +#include "cmemory.h" +#include "cstring.h" +#include "filestrm.h" +#include "toolutil.h" +#include "unicode/uclean.h" +#include "uoptions.h" +#include "pkg_genc.h" +#include "filetools.h" +#include "charstr.h" +#include "unicode/errorcode.h" + +#define MAX_COLUMN ((uint32_t)(0xFFFFFFFFU)) + +#define HEX_0X 0 /* 0x1234 */ +#define HEX_0H 1 /* 01234h */ + +/* prototypes --------------------------------------------------------------- */ +static void +getOutFilename( + const char *inFilename, + const char *destdir, + char *outFilename, + int32_t outFilenameCapacity, + char *entryName, + int32_t entryNameCapacity, + const char *newSuffix, + const char *optFilename); + +static uint32_t +write8(FileStream *out, uint8_t byte, uint32_t column); + +static uint32_t +write32(FileStream *out, uint32_t byte, uint32_t column); + +#if U_PLATFORM == U_PF_OS400 +static uint32_t +write8str(FileStream *out, uint8_t byte, uint32_t column); +#endif +/* -------------------------------------------------------------------------- */ + +/* +Creating Template Files for New Platforms + +Let the cc compiler help you get started. +Compile this program + const unsigned int x[5] = {1, 2, 0xdeadbeef, 0xffffffff, 16}; +with the -S option to produce assembly output. + +For example, this will generate array.s: +gcc -S array.c + +This will produce a .s file that may look like this: + + .file "array.c" + .version "01.01" +gcc2_compiled.: + .globl x + .section .rodata + .align 4 + .type x,@object + .size x,20 +x: + .long 1 + .long 2 + .long -559038737 + .long -1 + .long 16 + .ident "GCC: (GNU) 2.96 20000731 (Red Hat Linux 7.1 2.96-85)" + +which gives a starting point that will compile, and can be transformed +to become the template, generally with some consulting of as docs and +some experimentation. + +If you want ICU to automatically use this assembly, you should +specify "GENCCODE_ASSEMBLY=-a name" in the specific config/mh-* file, +where the name is the compiler or platform that you used in this +assemblyHeader data structure. +*/ +static const struct AssemblyType { + const char *name; + const char *header; + const char *beginLine; + const char *footer; + int8_t hexType; /* HEX_0X or HEX_0h */ +} assemblyHeader[] = { + /* For gcc assemblers, the meaning of .align changes depending on the */ + /* hardware, so we use .balign 16 which always means 16 bytes. */ + /* https://sourceware.org/binutils/docs/as/Pseudo-Ops.html */ + {"gcc", + ".globl %s\n" + "\t.section .note.GNU-stack,\"\",%%progbits\n" + "#ifdef __CET__\n" + "# include <cet.h>\n" + "#endif\n" + "\t.section .rodata\n" + "\t.balign 16\n" + "#ifdef U_HIDE_DATA_SYMBOL\n" + "\t.hidden %s\n" + "#endif\n" + "\t.type %s,%%object\n" + "%s:\n\n", + + ".long ",".size %s, .-%s\n",HEX_0X + }, + {"gcc-darwin", + /*"\t.section __TEXT,__text,regular,pure_instructions\n" + "\t.section __TEXT,__picsymbolstub1,symbol_stubs,pure_instructions,32\n"*/ + ".globl _%s\n" + "#ifdef U_HIDE_DATA_SYMBOL\n" + "\t.private_extern _%s\n" + "#endif\n" + "\t.data\n" + "\t.const\n" + "\t.balign 16\n" + "_%s:\n\n", + + ".long ","",HEX_0X + }, + /* macOS PPC should use `.p2align 4` instead `.balign 16` because is + * unknown pseudo ops for such legacy system*/ + {"gcc-darwin-ppc", + /*"\t.section __TEXT,__text,regular,pure_instructions\n" + "\t.section __TEXT,__picsymbolstub1,symbol_stubs,pure_instructions,32\n"*/ + ".globl _%s\n" + "#ifdef U_HIDE_DATA_SYMBOL\n" + "\t.private_extern _%s\n" + "#endif\n" + "\t.data\n" + "\t.const\n" + "\t.p2align 4\n" + "_%s:\n\n", + + ".long ","",HEX_0X + }, + {"gcc-cygwin", + ".globl _%s\n" + "\t.section .rodata\n" + "\t.balign 16\n" + "_%s:\n\n", + + ".long ","",HEX_0X + }, + {"gcc-mingw64", + ".globl %s\n" + "\t.section .rodata\n" + "\t.balign 16\n" + "%s:\n\n", + + ".long ","",HEX_0X + }, +/* 16 bytes alignment. */ +/* http://docs.oracle.com/cd/E19641-01/802-1947/802-1947.pdf */ + {"sun", + "\t.section \".rodata\"\n" + "\t.align 16\n" + ".globl %s\n" + "%s:\n", + + ".word ","",HEX_0X + }, +/* 16 bytes alignment for sun-x86. */ +/* http://docs.oracle.com/cd/E19963-01/html/821-1608/eoiyg.html */ + {"sun-x86", + "Drodata.rodata:\n" + "\t.type Drodata.rodata,@object\n" + "\t.size Drodata.rodata,0\n" + "\t.globl %s\n" + "\t.align 16\n" + "%s:\n", + + ".4byte ","",HEX_0X + }, +/* 1<<4 bit alignment for aix. */ +/* http://pic.dhe.ibm.com/infocenter/aix/v6r1/index.jsp?topic=%2Fcom.ibm.aix.aixassem%2Fdoc%2Falangref%2Fidalangref_csect_pseudoop.htm */ + {"xlc", + ".globl %s{RO}\n" + "\t.toc\n" + "%s:\n" + "\t.csect %s{RO}, 4\n", + + ".long ","",HEX_0X + }, + {"aCC-ia64", + "\t.file \"%s.s\"\n" + "\t.type %s,@object\n" + "\t.global %s\n" + "\t.secalias .abe$0.rodata, \".rodata\"\n" + "\t.section .abe$0.rodata = \"a\", \"progbits\"\n" + "\t.align 16\n" + "%s::\t", + + "data4 ","",HEX_0X + }, + {"aCC-parisc", + "\t.SPACE $TEXT$\n" + "\t.SUBSPA $LIT$\n" + "%s\n" + "\t.EXPORT %s\n" + "\t.ALIGN 16\n", + + ".WORD ","",HEX_0X + }, +/* align 16 bytes */ +/* http://msdn.microsoft.com/en-us/library/dwa9fwef.aspx */ + {"nasm", + "global %s\n" +#if defined(_WIN32) + "section .rdata align=16\n" +#else + "section .rodata align=16\n" +#endif + "%s:\n", + " dd ","",HEX_0X + }, + { "masm", + "\tTITLE %s\n" + "; generated by genccode\n" + ".386\n" + ".model flat\n" + "\tPUBLIC _%s\n" + "ICUDATA_%s\tSEGMENT READONLY PARA PUBLIC FLAT 'DATA'\n" + "\tALIGN 16\n" + "_%s\tLABEL DWORD\n", + "\tDWORD ","\nICUDATA_%s\tENDS\n\tEND\n",HEX_0H + }, + { "masm64", + "\tTITLE %s\n" + "; generated by genccode\n" + "\tPUBLIC _%s\n" + "ICUDATA_%s\tSEGMENT READONLY 'DATA'\n" + "\tALIGN 16\n" + "_%s\tLABEL DWORD\n", + "\tDWORD ","\nICUDATA_%s\tENDS\n\tEND\n",HEX_0H + } +}; + +static int32_t assemblyHeaderIndex = -1; +static int32_t hexType = HEX_0X; + +U_CAPI UBool U_EXPORT2 +checkAssemblyHeaderName(const char* optAssembly) { + int32_t idx; + assemblyHeaderIndex = -1; + for (idx = 0; idx < UPRV_LENGTHOF(assemblyHeader); idx++) { + if (uprv_strcmp(optAssembly, assemblyHeader[idx].name) == 0) { + assemblyHeaderIndex = idx; + hexType = assemblyHeader[idx].hexType; /* set the hex type */ + return true; + } + } + + return false; +} + + +U_CAPI void U_EXPORT2 +printAssemblyHeadersToStdErr() { + int32_t idx; + fprintf(stderr, "%s", assemblyHeader[0].name); + for (idx = 1; idx < UPRV_LENGTHOF(assemblyHeader); idx++) { + fprintf(stderr, ", %s", assemblyHeader[idx].name); + } + fprintf(stderr, + ")\n"); +} + +U_CAPI void U_EXPORT2 +writeAssemblyCode( + const char *filename, + const char *destdir, + const char *optEntryPoint, + const char *optFilename, + char *outFilePath, + size_t outFilePathCapacity) { + uint32_t column = MAX_COLUMN; + char entry[96]; + union { + uint32_t uint32s[1024]; + char chars[4096]; + } buffer; + FileStream *in, *out; + size_t i, length, count; + + in=T_FileStream_open(filename, "rb"); + if(in==nullptr) { + fprintf(stderr, "genccode: unable to open input file %s\n", filename); + exit(U_FILE_ACCESS_ERROR); + } + + const char* newSuffix = nullptr; + + if (uprv_strcmp(assemblyHeader[assemblyHeaderIndex].name, "masm") == 0) { + newSuffix = ".masm"; + } + else if (uprv_strcmp(assemblyHeader[assemblyHeaderIndex].name, "nasm") == 0) { + newSuffix = ".asm"; + } else { + newSuffix = ".S"; + } + + getOutFilename( + filename, + destdir, + buffer.chars, + sizeof(buffer.chars), + entry, + sizeof(entry), + newSuffix, + optFilename); + out=T_FileStream_open(buffer.chars, "w"); + if(out==nullptr) { + fprintf(stderr, "genccode: unable to open output file %s\n", buffer.chars); + exit(U_FILE_ACCESS_ERROR); + } + + if (outFilePath != nullptr) { + if (uprv_strlen(buffer.chars) >= outFilePathCapacity) { + fprintf(stderr, "genccode: filename too long\n"); + exit(U_ILLEGAL_ARGUMENT_ERROR); + } + uprv_strcpy(outFilePath, buffer.chars); +#if defined (WINDOWS_WITH_GNUC) && U_PLATFORM != U_PF_CYGWIN + /* Need to fix the file separator character when using MinGW. */ + swapFileSepChar(outFilePath, U_FILE_SEP_CHAR, '/'); +#endif + } + + if(optEntryPoint != nullptr) { + uprv_strcpy(entry, optEntryPoint); + uprv_strcat(entry, "_dat"); + } + + /* turn dashes or dots in the entry name into underscores */ + length=uprv_strlen(entry); + for(i=0; i<length; ++i) { + if(entry[i]=='-' || entry[i]=='.') { + entry[i]='_'; + } + } + + count = snprintf( + buffer.chars, sizeof(buffer.chars), + assemblyHeader[assemblyHeaderIndex].header, + entry, entry, entry, entry, + entry, entry, entry, entry); + if (count >= sizeof(buffer.chars)) { + fprintf(stderr, "genccode: entry name too long (long filename?)\n"); + exit(U_ILLEGAL_ARGUMENT_ERROR); + } + T_FileStream_writeLine(out, buffer.chars); + T_FileStream_writeLine(out, assemblyHeader[assemblyHeaderIndex].beginLine); + + for(;;) { + memset(buffer.uint32s, 0, sizeof(buffer.uint32s)); + length=T_FileStream_read(in, buffer.uint32s, sizeof(buffer.uint32s)); + if(length==0) { + break; + } + for(i=0; i<(length/sizeof(buffer.uint32s[0])); i++) { + // TODO: What if the last read sees length not as a multiple of 4? + column = write32(out, buffer.uint32s[i], column); + } + } + + T_FileStream_writeLine(out, "\n"); + + count = snprintf( + buffer.chars, sizeof(buffer.chars), + assemblyHeader[assemblyHeaderIndex].footer, + entry, entry, entry, entry, + entry, entry, entry, entry); + if (count >= sizeof(buffer.chars)) { + fprintf(stderr, "genccode: entry name too long (long filename?)\n"); + exit(U_ILLEGAL_ARGUMENT_ERROR); + } + T_FileStream_writeLine(out, buffer.chars); + + if(T_FileStream_error(in)) { + fprintf(stderr, "genccode: file read error while generating from file %s\n", filename); + exit(U_FILE_ACCESS_ERROR); + } + + if(T_FileStream_error(out)) { + fprintf(stderr, "genccode: file write error while generating from file %s\n", filename); + exit(U_FILE_ACCESS_ERROR); + } + + T_FileStream_close(out); + T_FileStream_close(in); +} + +U_CAPI void U_EXPORT2 +writeCCode( + const char *filename, + const char *destdir, + const char *optEntryPoint, + const char *optName, + const char *optFilename, + char *outFilePath, + size_t outFilePathCapacity) { + uint32_t column = MAX_COLUMN; + char buffer[4096], entry[96]; + FileStream *in, *out; + size_t i, length, count; + + in=T_FileStream_open(filename, "rb"); + if(in==nullptr) { + fprintf(stderr, "genccode: unable to open input file %s\n", filename); + exit(U_FILE_ACCESS_ERROR); + } + + if(optName != nullptr) { /* prepend 'icudt28_' */ + // +2 includes the _ and the NUL + if (uprv_strlen(optName) + 2 > sizeof(entry)) { + fprintf(stderr, "genccode: entry name too long (long filename?)\n"); + exit(U_ILLEGAL_ARGUMENT_ERROR); + } + strcpy(entry, optName); + strcat(entry, "_"); + } else { + entry[0] = 0; + } + + getOutFilename( + filename, + destdir, + buffer, + static_cast<int32_t>(sizeof(buffer)), + entry + uprv_strlen(entry), + static_cast<int32_t>(sizeof(entry) - uprv_strlen(entry)), + ".c", + optFilename); + + if (outFilePath != nullptr) { + if (uprv_strlen(buffer) >= outFilePathCapacity) { + fprintf(stderr, "genccode: filename too long\n"); + exit(U_ILLEGAL_ARGUMENT_ERROR); + } + uprv_strcpy(outFilePath, buffer); +#if defined (WINDOWS_WITH_GNUC) && U_PLATFORM != U_PF_CYGWIN + /* Need to fix the file separator character when using MinGW. */ + swapFileSepChar(outFilePath, U_FILE_SEP_CHAR, '/'); +#endif + } + + out=T_FileStream_open(buffer, "w"); + if(out==nullptr) { + fprintf(stderr, "genccode: unable to open output file %s\n", buffer); + exit(U_FILE_ACCESS_ERROR); + } + + if(optEntryPoint != nullptr) { + uprv_strcpy(entry, optEntryPoint); + uprv_strcat(entry, "_dat"); + } + + /* turn dashes or dots in the entry name into underscores */ + length=uprv_strlen(entry); + for(i=0; i<length; ++i) { + if(entry[i]=='-' || entry[i]=='.') { + entry[i]='_'; + } + } + +#if U_PLATFORM == U_PF_OS400 + /* + TODO: Fix this once the compiler implements this feature. Keep in sync with udatamem.c + + This is here because this platform can't currently put + const data into the read-only pages of an object or + shared library (service program). Only strings are allowed in read-only + pages, so we use char * strings to store the data. + + In order to prevent the beginning of the data from ever matching the + magic numbers we must still use the initial double. + [grhoten 4/24/2003] + */ + count = snprintf(buffer, sizeof(buffer), + "#ifndef IN_GENERATED_CCODE\n" + "#define IN_GENERATED_CCODE\n" + "#define U_DISABLE_RENAMING 1\n" + "#include \"unicode/umachine.h\"\n" + "#endif\n" + "U_CDECL_BEGIN\n" + "const struct {\n" + " double bogus;\n" + " const char *bytes; \n" + "} %s={ 0.0, \n", + entry); + if (count >= sizeof(buffer)) { + fprintf(stderr, "genccode: entry name too long (long filename?)\n"); + exit(U_ILLEGAL_ARGUMENT_ERROR); + } + T_FileStream_writeLine(out, buffer); + + for(;;) { + length=T_FileStream_read(in, buffer, sizeof(buffer)); + if(length==0) { + break; + } + for(i=0; i<length; ++i) { + column = write8str(out, (uint8_t)buffer[i], column); + } + } + + T_FileStream_writeLine(out, "\"\n};\nU_CDECL_END\n"); +#else + /* Function renaming shouldn't be done in data */ + count = snprintf(buffer, sizeof(buffer), + "#ifndef IN_GENERATED_CCODE\n" + "#define IN_GENERATED_CCODE\n" + "#define U_DISABLE_RENAMING 1\n" + "#include \"unicode/umachine.h\"\n" + "#endif\n" + "U_CDECL_BEGIN\n" + "const struct {\n" + " double bogus;\n" + " uint8_t bytes[%ld]; \n" + "} %s={ 0.0, {\n", + (long)T_FileStream_size(in), entry); + if (count >= sizeof(buffer)) { + fprintf(stderr, "genccode: entry name too long (long filename?)\n"); + exit(U_ILLEGAL_ARGUMENT_ERROR); + } + T_FileStream_writeLine(out, buffer); + + for(;;) { + length=T_FileStream_read(in, buffer, sizeof(buffer)); + if(length==0) { + break; + } + for(i=0; i<length; ++i) { + column = write8(out, (uint8_t)buffer[i], column); + } + } + + T_FileStream_writeLine(out, "\n}\n};\nU_CDECL_END\n"); +#endif + + if(T_FileStream_error(in)) { + fprintf(stderr, "genccode: file read error while generating from file %s\n", filename); + exit(U_FILE_ACCESS_ERROR); + } + + if(T_FileStream_error(out)) { + fprintf(stderr, "genccode: file write error while generating from file %s\n", filename); + exit(U_FILE_ACCESS_ERROR); + } + + T_FileStream_close(out); + T_FileStream_close(in); +} + +static uint32_t +write32(FileStream *out, uint32_t bitField, uint32_t column) { + int32_t i; + char bitFieldStr[64]; /* This is more bits than needed for a 32-bit number */ + char *s = bitFieldStr; + uint8_t *ptrIdx = (uint8_t *)&bitField; + static const char hexToStr[16] = { + '0','1','2','3', + '4','5','6','7', + '8','9','A','B', + 'C','D','E','F' + }; + + /* write the value, possibly with comma and newline */ + if(column==MAX_COLUMN) { + /* first byte */ + column=1; + } else if(column<32) { + *(s++)=','; + ++column; + } else { + *(s++)='\n'; + uprv_strcpy(s, assemblyHeader[assemblyHeaderIndex].beginLine); + s+=uprv_strlen(s); + column=1; + } + + if (bitField < 10) { + /* It's a small number. Don't waste the space for 0x */ + *(s++)=hexToStr[bitField]; + } + else { + int seenNonZero = 0; /* This is used to remove leading zeros */ + + if(hexType==HEX_0X) { + *(s++)='0'; + *(s++)='x'; + } else if(hexType==HEX_0H) { + *(s++)='0'; + } + + /* This creates a 32-bit field */ +#if U_IS_BIG_ENDIAN + for (i = 0; i < sizeof(uint32_t); i++) +#else + for (i = sizeof(uint32_t)-1; i >= 0 ; i--) +#endif + { + uint8_t value = ptrIdx[i]; + if (value || seenNonZero) { + *(s++)=hexToStr[value>>4]; + *(s++)=hexToStr[value&0xF]; + seenNonZero = 1; + } + } + if(hexType==HEX_0H) { + *(s++)='h'; + } + } + + *(s++)=0; + T_FileStream_writeLine(out, bitFieldStr); + return column; +} + +static uint32_t +write8(FileStream *out, uint8_t byte, uint32_t column) { + char s[4]; + int i=0; + + /* convert the byte value to a string */ + if(byte>=100) { + s[i++]=(char)('0'+byte/100); + byte%=100; + } + if(i>0 || byte>=10) { + s[i++]=(char)('0'+byte/10); + byte%=10; + } + s[i++]=(char)('0'+byte); + s[i]=0; + + /* write the value, possibly with comma and newline */ + if(column==MAX_COLUMN) { + /* first byte */ + column=1; + } else if(column<16) { + T_FileStream_writeLine(out, ","); + ++column; + } else { + T_FileStream_writeLine(out, ",\n"); + column=1; + } + T_FileStream_writeLine(out, s); + return column; +} + +#if U_PLATFORM == U_PF_OS400 +static uint32_t +write8str(FileStream *out, uint8_t byte, uint32_t column) { + char s[8]; + + if (byte > 7) + snprintf(s, sizeof(s), "\\x%X", byte); + else + snprintf(s, sizeof(s), "\\%X", byte); + + /* write the value, possibly with comma and newline */ + if(column==MAX_COLUMN) { + /* first byte */ + column=1; + T_FileStream_writeLine(out, "\""); + } else if(column<24) { + ++column; + } else { + T_FileStream_writeLine(out, "\"\n\""); + column=1; + } + T_FileStream_writeLine(out, s); + return column; +} +#endif + +static void +getOutFilename( + const char *inFilename, + const char *destdir, + char *outFilename, + int32_t outFilenameCapacity, + char *entryName, + int32_t entryNameCapacity, + const char *newSuffix, + const char *optFilename) { + const char *basename=findBasename(inFilename), *suffix=uprv_strrchr(basename, '.'); + + icu::CharString outFilenameBuilder; + icu::CharString entryNameBuilder; + icu::ErrorCode status; + + /* copy path */ + if(destdir!=nullptr && *destdir!=0) { + outFilenameBuilder.append(destdir, status); + outFilenameBuilder.ensureEndsWithFileSeparator(status); + } else { + outFilenameBuilder.append(inFilename, static_cast<int32_t>(basename - inFilename), status); + } + inFilename=basename; + + if(suffix==nullptr) { + /* the filename does not have a suffix */ + entryNameBuilder.append(inFilename, status); + if(optFilename != nullptr) { + outFilenameBuilder.append(optFilename, status); + } else { + outFilenameBuilder.append(inFilename, status); + } + outFilenameBuilder.append(newSuffix, status); + } else { + int32_t saveOutFilenameLength = outFilenameBuilder.length(); + /* copy basename */ + while(inFilename<suffix) { + // iSeries cannot have '-' in the .o objects. + char c = (*inFilename=='-') ? '_' : *inFilename; + outFilenameBuilder.append(c, status); + entryNameBuilder.append(c, status); + inFilename++; + } + + /* replace '.' by '_' */ + outFilenameBuilder.append('_', status); + entryNameBuilder.append('_', status); + ++inFilename; + + /* copy suffix */ + outFilenameBuilder.append(inFilename, status); + entryNameBuilder.append(inFilename, status); + + if(optFilename != nullptr) { + outFilenameBuilder.truncate(saveOutFilenameLength); + outFilenameBuilder.append(optFilename, status); + } + // add ".c" + outFilenameBuilder.append(newSuffix, status); + } + + if (status.isFailure()) { + fprintf(stderr, "genccode: error building filename or entrypoint\n"); + exit(status.get()); + } + + if (outFilenameBuilder.length() >= outFilenameCapacity) { + fprintf(stderr, "genccode: output filename too long\n"); + exit(U_ILLEGAL_ARGUMENT_ERROR); + } + + if (entryNameBuilder.length() >= entryNameCapacity) { + fprintf(stderr, "genccode: entry name too long (long filename?)\n"); + exit(U_ILLEGAL_ARGUMENT_ERROR); + } + + outFilenameBuilder.extract(outFilename, outFilenameCapacity, status); + entryNameBuilder.extract(entryName, entryNameCapacity, status); +} + +#ifdef CAN_GENERATE_OBJECTS +static void +getArchitecture(uint16_t *pCPU, uint16_t *pBits, UBool *pIsBigEndian, const char *optMatchArch) { + union { + char bytes[2048]; +#ifdef U_ELF + Elf32_Ehdr header32; + /* Elf32_Ehdr and ELF64_Ehdr are identical for the necessary fields. */ +#elif U_PLATFORM_HAS_WIN32_API + IMAGE_FILE_HEADER header; +#endif + } buffer; + + const char *filename; + FileStream *in; + int32_t length; + +#ifdef U_ELF + +#elif U_PLATFORM_HAS_WIN32_API + const IMAGE_FILE_HEADER *pHeader; +#else +# error "Unknown platform for CAN_GENERATE_OBJECTS." +#endif + + if(optMatchArch != nullptr) { + filename=optMatchArch; + } else { + /* set defaults */ +#ifdef U_ELF + /* set EM_386 because elf.h does not provide better defaults */ + *pCPU=EM_386; + *pBits=32; + *pIsBigEndian=(UBool)(U_IS_BIG_ENDIAN ? ELFDATA2MSB : ELFDATA2LSB); +#elif U_PLATFORM_HAS_WIN32_API + // Windows always runs in little-endian mode. + *pIsBigEndian = false; + + // Note: The various _M_<arch> macros are predefined by the MSVC compiler based + // on the target compilation architecture. + // https://docs.microsoft.com/cpp/preprocessor/predefined-macros + + // link.exe will link an IMAGE_FILE_MACHINE_UNKNOWN data-only .obj file + // no matter what architecture it is targeting (though other values are + // required to match). Unfortunately, the variable name decoration/mangling + // is slightly different on x86, which means we can't use the UNKNOWN type + // for all architectures though. +# if defined(_M_IX86) + *pCPU = IMAGE_FILE_MACHINE_I386; +# else + *pCPU = IMAGE_FILE_MACHINE_UNKNOWN; +# endif +# if defined(_M_IA64) || defined(_M_AMD64) || defined (_M_ARM64) + *pBits = 64; // Doesn't seem to be used for anything interesting though? +# elif defined(_M_IX86) || defined(_M_ARM) + *pBits = 32; +# else +# error "Unknown platform for CAN_GENERATE_OBJECTS." +# endif +#else +# error "Unknown platform for CAN_GENERATE_OBJECTS." +#endif + return; + } + + in=T_FileStream_open(filename, "rb"); + if(in==nullptr) { + fprintf(stderr, "genccode: unable to open match-arch file %s\n", filename); + exit(U_FILE_ACCESS_ERROR); + } + length=T_FileStream_read(in, buffer.bytes, sizeof(buffer.bytes)); + +#ifdef U_ELF + if(length<(int32_t)sizeof(Elf32_Ehdr)) { + fprintf(stderr, "genccode: match-arch file %s is too short\n", filename); + exit(U_UNSUPPORTED_ERROR); + } + if( + buffer.header32.e_ident[0]!=ELFMAG0 || + buffer.header32.e_ident[1]!=ELFMAG1 || + buffer.header32.e_ident[2]!=ELFMAG2 || + buffer.header32.e_ident[3]!=ELFMAG3 || + buffer.header32.e_ident[EI_CLASS]<ELFCLASS32 || buffer.header32.e_ident[EI_CLASS]>ELFCLASS64 + ) { + fprintf(stderr, "genccode: match-arch file %s is not an ELF object file, or not supported\n", filename); + exit(U_UNSUPPORTED_ERROR); + } + + *pBits= buffer.header32.e_ident[EI_CLASS]==ELFCLASS32 ? 32 : 64; /* only 32 or 64: see check above */ +#ifdef U_ELF64 + if(*pBits!=32 && *pBits!=64) { + fprintf(stderr, "genccode: currently only supports 32-bit and 64-bit ELF format\n"); + exit(U_UNSUPPORTED_ERROR); + } +#else + if(*pBits!=32) { + fprintf(stderr, "genccode: built with elf.h missing 64-bit definitions\n"); + exit(U_UNSUPPORTED_ERROR); + } +#endif + + *pIsBigEndian=(UBool)(buffer.header32.e_ident[EI_DATA]==ELFDATA2MSB); + if(*pIsBigEndian!=U_IS_BIG_ENDIAN) { + fprintf(stderr, "genccode: currently only same-endianness ELF formats are supported\n"); + exit(U_UNSUPPORTED_ERROR); + } + /* TODO: Support byte swapping */ + + *pCPU=buffer.header32.e_machine; +#elif U_PLATFORM_HAS_WIN32_API + if(length<sizeof(IMAGE_FILE_HEADER)) { + fprintf(stderr, "genccode: match-arch file %s is too short\n", filename); + exit(U_UNSUPPORTED_ERROR); + } + /* TODO: Use buffer.header. Keep aliasing legal. */ + pHeader=(const IMAGE_FILE_HEADER *)buffer.bytes; + *pCPU=pHeader->Machine; + /* + * The number of bits is implicit with the Machine value. + * *pBits is ignored in the calling code, so this need not be precise. + */ + *pBits= *pCPU==IMAGE_FILE_MACHINE_I386 ? 32 : 64; + /* Windows always runs on little-endian CPUs. */ + *pIsBigEndian=false; +#else +# error "Unknown platform for CAN_GENERATE_OBJECTS." +#endif + + T_FileStream_close(in); +} + +U_CAPI void U_EXPORT2 +writeObjectCode( + const char *filename, + const char *destdir, + const char *optEntryPoint, + const char *optMatchArch, + const char *optFilename, + char *outFilePath, + size_t outFilePathCapacity, + UBool optWinDllExport) { + /* common variables */ + char buffer[4096], entry[96]={ 0 }; + FileStream *in, *out; + const char *newSuffix; + int32_t i, entryLength, length, size, entryOffset=0, entryLengthOffset=0; + + uint16_t cpu, bits; + UBool makeBigEndian; + + (void)optWinDllExport; /* unused except Windows */ + + /* platform-specific variables and initialization code */ +#ifdef U_ELF + /* 32-bit Elf file header */ + static Elf32_Ehdr header32={ + { + /* e_ident[] */ + ELFMAG0, ELFMAG1, ELFMAG2, ELFMAG3, + ELFCLASS32, + U_IS_BIG_ENDIAN ? ELFDATA2MSB : ELFDATA2LSB, + EV_CURRENT /* EI_VERSION */ + }, + ET_REL, + EM_386, + EV_CURRENT, /* e_version */ + 0, /* e_entry */ + 0, /* e_phoff */ + (Elf32_Off)sizeof(Elf32_Ehdr), /* e_shoff */ + 0, /* e_flags */ + (Elf32_Half)sizeof(Elf32_Ehdr), /* eh_size */ + 0, /* e_phentsize */ + 0, /* e_phnum */ + (Elf32_Half)sizeof(Elf32_Shdr), /* e_shentsize */ + 5, /* e_shnum */ + 2 /* e_shstrndx */ + }; + + /* 32-bit Elf section header table */ + static Elf32_Shdr sectionHeaders32[5]={ + { /* SHN_UNDEF */ + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 + }, + { /* .symtab */ + 1, /* sh_name */ + SHT_SYMTAB, + 0, /* sh_flags */ + 0, /* sh_addr */ + (Elf32_Off)(sizeof(header32)+sizeof(sectionHeaders32)), /* sh_offset */ + (Elf32_Word)(2*sizeof(Elf32_Sym)), /* sh_size */ + 3, /* sh_link=sect hdr index of .strtab */ + 1, /* sh_info=One greater than the symbol table index of the last + * local symbol (with STB_LOCAL). */ + 4, /* sh_addralign */ + (Elf32_Word)(sizeof(Elf32_Sym)) /* sh_entsize */ + }, + { /* .shstrtab */ + 9, /* sh_name */ + SHT_STRTAB, + 0, /* sh_flags */ + 0, /* sh_addr */ + (Elf32_Off)(sizeof(header32)+sizeof(sectionHeaders32)+2*sizeof(Elf32_Sym)), /* sh_offset */ + 40, /* sh_size */ + 0, /* sh_link */ + 0, /* sh_info */ + 1, /* sh_addralign */ + 0 /* sh_entsize */ + }, + { /* .strtab */ + 19, /* sh_name */ + SHT_STRTAB, + 0, /* sh_flags */ + 0, /* sh_addr */ + (Elf32_Off)(sizeof(header32)+sizeof(sectionHeaders32)+2*sizeof(Elf32_Sym)+40), /* sh_offset */ + (Elf32_Word)sizeof(entry), /* sh_size */ + 0, /* sh_link */ + 0, /* sh_info */ + 1, /* sh_addralign */ + 0 /* sh_entsize */ + }, + { /* .rodata */ + 27, /* sh_name */ + SHT_PROGBITS, + SHF_ALLOC, /* sh_flags */ + 0, /* sh_addr */ + (Elf32_Off)(sizeof(header32)+sizeof(sectionHeaders32)+2*sizeof(Elf32_Sym)+40+sizeof(entry)), /* sh_offset */ + 0, /* sh_size */ + 0, /* sh_link */ + 0, /* sh_info */ + 16, /* sh_addralign */ + 0 /* sh_entsize */ + } + }; + + /* symbol table */ + static Elf32_Sym symbols32[2]={ + { /* STN_UNDEF */ + 0, 0, 0, 0, 0, 0 + }, + { /* data entry point */ + 1, /* st_name */ + 0, /* st_value */ + 0, /* st_size */ + ELF64_ST_INFO(STB_GLOBAL, STT_OBJECT), + 0, /* st_other */ + 4 /* st_shndx=index of related section table entry */ + } + }; + + /* section header string table, with decimal string offsets */ + static const char sectionStrings[40]= + /* 0 */ "\0" + /* 1 */ ".symtab\0" + /* 9 */ ".shstrtab\0" + /* 19 */ ".strtab\0" + /* 27 */ ".rodata\0" + /* 35 */ "\0\0\0\0"; /* contains terminating NUL */ + /* 40: padded to multiple of 8 bytes */ + + /* + * Use entry[] for the string table which will contain only the + * entry point name. + * entry[0] must be 0 (NUL) + * The entry point name can be up to 38 characters long (sizeof(entry)-2). + */ + + /* 16-align .rodata in the .o file, just in case */ + static const char padding[16]={ 0 }; + int32_t paddingSize; + +#ifdef U_ELF64 + /* 64-bit Elf file header */ + static Elf64_Ehdr header64={ + { + /* e_ident[] */ + ELFMAG0, ELFMAG1, ELFMAG2, ELFMAG3, + ELFCLASS64, + U_IS_BIG_ENDIAN ? ELFDATA2MSB : ELFDATA2LSB, + EV_CURRENT /* EI_VERSION */ + }, + ET_REL, + EM_X86_64, + EV_CURRENT, /* e_version */ + 0, /* e_entry */ + 0, /* e_phoff */ + (Elf64_Off)sizeof(Elf64_Ehdr), /* e_shoff */ + 0, /* e_flags */ + (Elf64_Half)sizeof(Elf64_Ehdr), /* eh_size */ + 0, /* e_phentsize */ + 0, /* e_phnum */ + (Elf64_Half)sizeof(Elf64_Shdr), /* e_shentsize */ + 5, /* e_shnum */ + 2 /* e_shstrndx */ + }; + + /* 64-bit Elf section header table */ + static Elf64_Shdr sectionHeaders64[5]={ + { /* SHN_UNDEF */ + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 + }, + { /* .symtab */ + 1, /* sh_name */ + SHT_SYMTAB, + 0, /* sh_flags */ + 0, /* sh_addr */ + (Elf64_Off)(sizeof(header64)+sizeof(sectionHeaders64)), /* sh_offset */ + (Elf64_Xword)(2*sizeof(Elf64_Sym)), /* sh_size */ + 3, /* sh_link=sect hdr index of .strtab */ + 1, /* sh_info=One greater than the symbol table index of the last + * local symbol (with STB_LOCAL). */ + 4, /* sh_addralign */ + (Elf64_Xword)(sizeof(Elf64_Sym)) /* sh_entsize */ + }, + { /* .shstrtab */ + 9, /* sh_name */ + SHT_STRTAB, + 0, /* sh_flags */ + 0, /* sh_addr */ + (Elf64_Off)(sizeof(header64)+sizeof(sectionHeaders64)+2*sizeof(Elf64_Sym)), /* sh_offset */ + 40, /* sh_size */ + 0, /* sh_link */ + 0, /* sh_info */ + 1, /* sh_addralign */ + 0 /* sh_entsize */ + }, + { /* .strtab */ + 19, /* sh_name */ + SHT_STRTAB, + 0, /* sh_flags */ + 0, /* sh_addr */ + (Elf64_Off)(sizeof(header64)+sizeof(sectionHeaders64)+2*sizeof(Elf64_Sym)+40), /* sh_offset */ + (Elf64_Xword)sizeof(entry), /* sh_size */ + 0, /* sh_link */ + 0, /* sh_info */ + 1, /* sh_addralign */ + 0 /* sh_entsize */ + }, + { /* .rodata */ + 27, /* sh_name */ + SHT_PROGBITS, + SHF_ALLOC, /* sh_flags */ + 0, /* sh_addr */ + (Elf64_Off)(sizeof(header64)+sizeof(sectionHeaders64)+2*sizeof(Elf64_Sym)+40+sizeof(entry)), /* sh_offset */ + 0, /* sh_size */ + 0, /* sh_link */ + 0, /* sh_info */ + 16, /* sh_addralign */ + 0 /* sh_entsize */ + } + }; + + /* + * 64-bit symbol table + * careful: different order of items compared with Elf32_sym! + */ + static Elf64_Sym symbols64[2]={ + { /* STN_UNDEF */ + 0, 0, 0, 0, 0, 0 + }, + { /* data entry point */ + 1, /* st_name */ + ELF64_ST_INFO(STB_GLOBAL, STT_OBJECT), + 0, /* st_other */ + 4, /* st_shndx=index of related section table entry */ + 0, /* st_value */ + 0 /* st_size */ + } + }; + +#endif /* U_ELF64 */ + + /* entry[] have a leading NUL */ + entryOffset=1; + + /* in the common code, count entryLength from after the NUL */ + entryLengthOffset=1; + + newSuffix=".o"; + +#elif U_PLATFORM_HAS_WIN32_API + struct { + IMAGE_FILE_HEADER fileHeader; + IMAGE_SECTION_HEADER sections[2]; + char linkerOptions[100]; + } objHeader; + IMAGE_SYMBOL symbols[1]; + struct { + DWORD sizeofLongNames; + char longNames[100]; + } symbolNames; + + /* + * entry sometimes have a leading '_' + * overwritten if entryOffset==0 depending on the target platform + * see check for cpu below + */ + entry[0]='_'; + + newSuffix=".obj"; +#else +# error "Unknown platform for CAN_GENERATE_OBJECTS." +#endif + + /* deal with options, files and the entry point name */ + getArchitecture(&cpu, &bits, &makeBigEndian, optMatchArch); + if (optMatchArch) + { + printf("genccode: --match-arch cpu=%hu bits=%hu big-endian=%d\n", cpu, bits, makeBigEndian); + } + else + { + printf("genccode: using architecture cpu=%hu bits=%hu big-endian=%d\n", cpu, bits, makeBigEndian); + } +#if U_PLATFORM_HAS_WIN32_API + if(cpu==IMAGE_FILE_MACHINE_I386) { + entryOffset=1; + } +#endif + + in=T_FileStream_open(filename, "rb"); + if(in==nullptr) { + fprintf(stderr, "genccode: unable to open input file %s\n", filename); + exit(U_FILE_ACCESS_ERROR); + } + size=T_FileStream_size(in); + + getOutFilename( + filename, + destdir, + buffer, + sizeof(buffer), + entry + entryOffset, + sizeof(entry) - entryOffset, + newSuffix, + optFilename); + + if (outFilePath != nullptr) { + if (uprv_strlen(buffer) >= outFilePathCapacity) { + fprintf(stderr, "genccode: filename too long\n"); + exit(U_ILLEGAL_ARGUMENT_ERROR); + } + uprv_strcpy(outFilePath, buffer); + } + + if(optEntryPoint != nullptr) { + uprv_strcpy(entry+entryOffset, optEntryPoint); + uprv_strcat(entry+entryOffset, "_dat"); + } + /* turn dashes in the entry name into underscores */ + entryLength=(int32_t)uprv_strlen(entry+entryLengthOffset); + for(i=0; i<entryLength; ++i) { + if(entry[entryLengthOffset+i]=='-') { + entry[entryLengthOffset+i]='_'; + } + } + + /* open the output file */ + out=T_FileStream_open(buffer, "wb"); + if(out==nullptr) { + fprintf(stderr, "genccode: unable to open output file %s\n", buffer); + exit(U_FILE_ACCESS_ERROR); + } + +#ifdef U_ELF + if(bits==32) { + header32.e_ident[EI_DATA]= makeBigEndian ? ELFDATA2MSB : ELFDATA2LSB; + header32.e_machine=cpu; + + /* 16-align .rodata in the .o file, just in case */ + paddingSize=sectionHeaders32[4].sh_offset & 0xf; + if(paddingSize!=0) { + paddingSize=0x10-paddingSize; + sectionHeaders32[4].sh_offset+=paddingSize; + } + + sectionHeaders32[4].sh_size=(Elf32_Word)size; + + symbols32[1].st_size=(Elf32_Word)size; + + /* write .o headers */ + T_FileStream_write(out, &header32, (int32_t)sizeof(header32)); + T_FileStream_write(out, sectionHeaders32, (int32_t)sizeof(sectionHeaders32)); + T_FileStream_write(out, symbols32, (int32_t)sizeof(symbols32)); + } else /* bits==64 */ { +#ifdef U_ELF64 + header64.e_ident[EI_DATA]= makeBigEndian ? ELFDATA2MSB : ELFDATA2LSB; + header64.e_machine=cpu; + + /* 16-align .rodata in the .o file, just in case */ + paddingSize=sectionHeaders64[4].sh_offset & 0xf; + if(paddingSize!=0) { + paddingSize=0x10-paddingSize; + sectionHeaders64[4].sh_offset+=paddingSize; + } + + sectionHeaders64[4].sh_size=(Elf64_Xword)size; + + symbols64[1].st_size=(Elf64_Xword)size; + + /* write .o headers */ + T_FileStream_write(out, &header64, (int32_t)sizeof(header64)); + T_FileStream_write(out, sectionHeaders64, (int32_t)sizeof(sectionHeaders64)); + T_FileStream_write(out, symbols64, (int32_t)sizeof(symbols64)); +#endif + } + + T_FileStream_write(out, sectionStrings, (int32_t)sizeof(sectionStrings)); + T_FileStream_write(out, entry, (int32_t)sizeof(entry)); + if(paddingSize!=0) { + T_FileStream_write(out, padding, paddingSize); + } +#elif U_PLATFORM_HAS_WIN32_API + /* populate the .obj headers */ + uprv_memset(&objHeader, 0, sizeof(objHeader)); + uprv_memset(&symbols, 0, sizeof(symbols)); + uprv_memset(&symbolNames, 0, sizeof(symbolNames)); + + /* write the linker export directive */ + if (optWinDllExport) { + uprv_strcpy(objHeader.linkerOptions, "-export:"); + length=8; + uprv_strcpy(objHeader.linkerOptions+length, entry); + length+=entryLength; + uprv_strcpy(objHeader.linkerOptions+length, ",data "); + length+=6; + } + else { + length=0; + } + + /* set the file header */ + objHeader.fileHeader.Machine=cpu; + objHeader.fileHeader.NumberOfSections=2; + objHeader.fileHeader.TimeDateStamp=(DWORD)time(nullptr); + objHeader.fileHeader.PointerToSymbolTable=IMAGE_SIZEOF_FILE_HEADER+2*IMAGE_SIZEOF_SECTION_HEADER+length+size; /* start of symbol table */ + objHeader.fileHeader.NumberOfSymbols=1; + + /* set the section for the linker options */ + uprv_strncpy((char *)objHeader.sections[0].Name, ".drectve", 8); + objHeader.sections[0].SizeOfRawData=length; + objHeader.sections[0].PointerToRawData=IMAGE_SIZEOF_FILE_HEADER+2*IMAGE_SIZEOF_SECTION_HEADER; + objHeader.sections[0].Characteristics=IMAGE_SCN_LNK_INFO|IMAGE_SCN_LNK_REMOVE|IMAGE_SCN_ALIGN_1BYTES; + + /* set the data section */ + uprv_strncpy((char *)objHeader.sections[1].Name, ".rdata", 6); + objHeader.sections[1].SizeOfRawData=size; + objHeader.sections[1].PointerToRawData=IMAGE_SIZEOF_FILE_HEADER+2*IMAGE_SIZEOF_SECTION_HEADER+length; + objHeader.sections[1].Characteristics=IMAGE_SCN_CNT_INITIALIZED_DATA|IMAGE_SCN_ALIGN_16BYTES|IMAGE_SCN_MEM_READ; + + /* set the symbol table */ + if(entryLength<=8) { + uprv_strncpy((char *)symbols[0].N.ShortName, entry, entryLength); + symbolNames.sizeofLongNames=4; + } else { + symbols[0].N.Name.Short=0; + symbols[0].N.Name.Long=4; + symbolNames.sizeofLongNames=4+entryLength+1; + uprv_strcpy(symbolNames.longNames, entry); + } + symbols[0].SectionNumber=2; + symbols[0].StorageClass=IMAGE_SYM_CLASS_EXTERNAL; + + /* write the file header and the linker options section */ + T_FileStream_write(out, &objHeader, objHeader.sections[1].PointerToRawData); +#else +# error "Unknown platform for CAN_GENERATE_OBJECTS." +#endif + + /* copy the data file into section 2 */ + for(;;) { + length=T_FileStream_read(in, buffer, sizeof(buffer)); + if(length==0) { + break; + } + T_FileStream_write(out, buffer, (int32_t)length); + } + +#if U_PLATFORM_HAS_WIN32_API + /* write the symbol table */ + T_FileStream_write(out, symbols, IMAGE_SIZEOF_SYMBOL); + T_FileStream_write(out, &symbolNames, symbolNames.sizeofLongNames); +#endif + + if(T_FileStream_error(in)) { + fprintf(stderr, "genccode: file read error while generating from file %s\n", filename); + exit(U_FILE_ACCESS_ERROR); + } + + if(T_FileStream_error(out)) { + fprintf(stderr, "genccode: file write error while generating from file %s\n", filename); + exit(U_FILE_ACCESS_ERROR); + } + + T_FileStream_close(out); + T_FileStream_close(in); +} +#endif |