diff options
author | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-07 16:11:47 +0000 |
---|---|---|
committer | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-07 16:11:47 +0000 |
commit | 758f820bcc0f68aeebac1717e537ca13a320b909 (patch) | |
tree | 48111ece75cf4f98316848b37a7e26356e00669e /lib/exclude.c | |
parent | Initial commit. (diff) | |
download | coreutils-upstream.tar.xz coreutils-upstream.zip |
Adding upstream version 9.1.upstream/9.1upstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'lib/exclude.c')
-rw-r--r-- | lib/exclude.c | 689 |
1 files changed, 689 insertions, 0 deletions
diff --git a/lib/exclude.c b/lib/exclude.c new file mode 100644 index 0000000..3770c48 --- /dev/null +++ b/lib/exclude.c @@ -0,0 +1,689 @@ +/* exclude.c -- exclude file names + + Copyright (C) 1992-1994, 1997, 1999-2007, 2009-2022 Free Software + Foundation, Inc. + + 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 3 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, see <https://www.gnu.org/licenses/>. */ + +/* Written by Paul Eggert <eggert@twinsun.com> + and Sergey Poznyakoff <gray@gnu.org>. + Thanks to Phil Proudman <phil@proudman51.freeserve.co.uk> + for improvement suggestions. */ + +#include <config.h> + +#include <stdbool.h> + +#include <ctype.h> +#include <errno.h> +#include <stddef.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <wctype.h> +#include <regex.h> + +#include "exclude.h" +#include "hash.h" +#include "mbuiter.h" +#include "fnmatch.h" +#include "xalloc.h" +#include "verify.h" +#include "filename.h" + +#if GNULIB_EXCLUDE_SINGLE_THREAD +# include "unlocked-io.h" +#endif + +/* Non-GNU systems lack these options, so we don't need to check them. */ +#ifndef FNM_CASEFOLD +# define FNM_CASEFOLD 0 +#endif +#ifndef FNM_EXTMATCH +# define FNM_EXTMATCH 0 +#endif +#ifndef FNM_LEADING_DIR +# define FNM_LEADING_DIR 0 +#endif + +verify (((EXCLUDE_ANCHORED | EXCLUDE_INCLUDE | EXCLUDE_WILDCARDS) + & (FNM_PATHNAME | FNM_NOESCAPE | FNM_PERIOD | FNM_LEADING_DIR + | FNM_CASEFOLD | FNM_EXTMATCH)) + == 0); + + +/* Exclusion patterns are grouped into a singly-linked list of + "exclusion segments". Each segment represents a set of patterns + that can be matches using the same algorithm. Non-wildcard + patterns are kept in hash tables, to speed up searches. Wildcard + patterns are stored as arrays of patterns. */ + + +/* An exclude pattern-options pair. The options are fnmatch options + ORed with EXCLUDE_* options. */ + +struct patopts + { + int options; + union + { + char const *pattern; + regex_t re; + } v; + }; + +/* An array of pattern-options pairs. */ + +struct exclude_pattern + { + struct patopts *exclude; + idx_t exclude_alloc; + idx_t exclude_count; + }; + +enum exclude_type + { + exclude_hash, /* a hash table of excluded names */ + exclude_pattern /* an array of exclude patterns */ + }; + +struct exclude_segment + { + struct exclude_segment *next; /* next segment in list */ + enum exclude_type type; /* type of this segment */ + int options; /* common options for this segment */ + union + { + Hash_table *table; /* for type == exclude_hash */ + struct exclude_pattern pat; /* for type == exclude_pattern */ + } v; + }; + +struct pattern_buffer + { + struct pattern_buffer *next; + char *base; + }; + +/* The exclude structure keeps a singly-linked list of exclude segments, + maintained in reverse order. */ +struct exclude + { + struct exclude_segment *head; + struct pattern_buffer *patbuf; + }; + +/* Register BUF in the pattern buffer list of EX. ADD_FUNC (see + add_exclude_file and add_exclude_fp below) can use this function + if it modifies the pattern, to ensure the allocated memory will be + properly reclaimed upon calling free_exclude. */ +void +exclude_add_pattern_buffer (struct exclude *ex, char *buf) +{ + struct pattern_buffer *pbuf = xmalloc (sizeof *pbuf); + pbuf->base = buf; + pbuf->next = ex->patbuf; + ex->patbuf = pbuf; +} + +/* Return true if STR has or may have wildcards, when matched with OPTIONS. + Return false if STR definitely does not have wildcards. */ +bool +fnmatch_pattern_has_wildcards (const char *str, int options) +{ + while (1) + { + switch (*str++) + { + case '.': + case '{': + case '}': + case '(': + case ')': + if (options & EXCLUDE_REGEX) + return true; + break; + + case '\\': + if (options & EXCLUDE_REGEX) + continue; + else + str += ! (options & FNM_NOESCAPE) && *str; + break; + + case '+': case '@': case '!': + if (options & FNM_EXTMATCH && *str == '(') + return true; + break; + + case '?': case '*': case '[': + return true; + + case '\0': + return false; + } + } +} + +static void +unescape_pattern (char *str) +{ + char const *q = str; + do + q += *q == '\\' && q[1]; + while ((*str++ = *q++)); +} + +/* Return a newly allocated and empty exclude list. */ + +struct exclude * +new_exclude (void) +{ + return xzalloc (sizeof *new_exclude ()); +} + +/* Calculate the hash of string. */ +static size_t +string_hasher (void const *data, size_t n_buckets) +{ + char const *p = data; + return hash_string (p, n_buckets); +} + +/* Ditto, for case-insensitive hashes */ +static size_t +string_hasher_ci (void const *data, size_t n_buckets) +{ + char const *p = data; + mbui_iterator_t iter; + size_t value = 0; + + for (mbui_init (iter, p); mbui_avail (iter); mbui_advance (iter)) + { + mbchar_t m = mbui_cur (iter); + wchar_t wc; + + if (m.wc_valid) + wc = towlower (m.wc); + else + wc = *m.ptr; + + value = value * 31 + wc; + } + + return value % n_buckets; +} + +/* compare two strings for equality */ +static bool +string_compare (void const *data1, void const *data2) +{ + char const *p1 = data1; + char const *p2 = data2; + return strcmp (p1, p2) == 0; +} + +/* compare two strings for equality, case-insensitive */ +static bool +string_compare_ci (void const *data1, void const *data2) +{ + char const *p1 = data1; + char const *p2 = data2; + return mbscasecmp (p1, p2) == 0; +} + +static void +string_free (void *data) +{ + free (data); +} + +/* Create new exclude segment of given TYPE and OPTIONS, and attach it + to the head of EX. */ +static void +new_exclude_segment (struct exclude *ex, enum exclude_type type, int options) +{ + struct exclude_segment *sp = xzalloc (sizeof (struct exclude_segment)); + sp->type = type; + sp->options = options; + switch (type) + { + case exclude_pattern: + break; + + case exclude_hash: + sp->v.table = hash_initialize (0, NULL, + (options & FNM_CASEFOLD) ? + string_hasher_ci + : string_hasher, + (options & FNM_CASEFOLD) ? + string_compare_ci + : string_compare, + string_free); + break; + } + sp->next = ex->head; + ex->head = sp; +} + +/* Free a single exclude segment */ +static void +free_exclude_segment (struct exclude_segment *seg) +{ + switch (seg->type) + { + case exclude_pattern: + for (idx_t i = 0; i < seg->v.pat.exclude_count; i++) + { + if (seg->v.pat.exclude[i].options & EXCLUDE_REGEX) + regfree (&seg->v.pat.exclude[i].v.re); + } + free (seg->v.pat.exclude); + break; + + case exclude_hash: + hash_free (seg->v.table); + break; + } + free (seg); +} + +/* Free the storage associated with an exclude list. */ +void +free_exclude (struct exclude *ex) +{ + struct exclude_segment *seg; + struct pattern_buffer *pbuf; + + for (seg = ex->head; seg; ) + { + struct exclude_segment *next = seg->next; + free_exclude_segment (seg); + seg = next; + } + + for (pbuf = ex->patbuf; pbuf; ) + { + struct pattern_buffer *next = pbuf->next; + free (pbuf->base); + free (pbuf); + pbuf = next; + } + + free (ex); +} + +/* Return zero if PATTERN matches F, obeying OPTIONS, except that + (unlike fnmatch) wildcards are disabled in PATTERN. */ + +static int +fnmatch_no_wildcards (char const *pattern, char const *f, int options) +{ + if (! (options & FNM_LEADING_DIR)) + return ((options & FNM_CASEFOLD) + ? mbscasecmp (pattern, f) + : strcmp (pattern, f)); + else if (! (options & FNM_CASEFOLD)) + { + size_t patlen = strlen (pattern); + int r = strncmp (pattern, f, patlen); + if (! r) + { + r = f[patlen]; + if (r == '/') + r = 0; + } + return r; + } + else + { + /* Walk through a copy of F, seeing whether P matches any prefix + of F. + + FIXME: This is an O(N**2) algorithm; it should be O(N). + Also, the copy should not be necessary. However, fixing this + will probably involve a change to the mbs* API. */ + + char *fcopy = xstrdup (f); + char *p; + int r; + for (p = fcopy; ; *p++ = '/') + { + p = strchr (p, '/'); + if (p) + *p = '\0'; + r = mbscasecmp (pattern, fcopy); + if (!p || r <= 0) + break; + } + free (fcopy); + return r; + } +} + +bool +exclude_fnmatch (char const *pattern, char const *f, int options) +{ + int (*matcher) (char const *, char const *, int) = + (options & EXCLUDE_WILDCARDS + ? fnmatch + : fnmatch_no_wildcards); + bool matched = ((*matcher) (pattern, f, options) == 0); + char const *p; + + if (! (options & EXCLUDE_ANCHORED)) + for (p = f; *p && ! matched; p++) + if (*p == '/' && p[1] != '/') + matched = ((*matcher) (pattern, p + 1, options) == 0); + + return matched; +} + +static bool +exclude_patopts (struct patopts const *opts, char const *f) +{ + int options = opts->options; + + return (options & EXCLUDE_REGEX) + ? regexec (&opts->v.re, f, 0, NULL, 0) == 0 + : exclude_fnmatch (opts->v.pattern, f, options); +} + +/* Return true if the exclude_pattern segment SEG matches F. */ + +static bool +file_pattern_matches (struct exclude_segment const *seg, char const *f) +{ + idx_t exclude_count = seg->v.pat.exclude_count; + struct patopts const *exclude = seg->v.pat.exclude; + + for (idx_t i = 0; i < exclude_count; i++) + { + if (exclude_patopts (exclude + i, f)) + return true; + } + return false; +} + +/* Return true if the exclude_hash segment SEG matches F. + BUFFER is an auxiliary storage of the same length as F (with nul + terminator included) */ +static bool +file_name_matches (struct exclude_segment const *seg, char const *f, + char *buffer) +{ + int options = seg->options; + Hash_table *table = seg->v.table; + + do + { + /* initialize the pattern */ + strcpy (buffer, f); + + while (1) + { + if (hash_lookup (table, buffer)) + return true; + if (options & FNM_LEADING_DIR) + { + char *p = strrchr (buffer, '/'); + if (p) + { + *p = 0; + continue; + } + } + break; + } + + if (!(options & EXCLUDE_ANCHORED)) + { + f = strchr (f, '/'); + if (f) + f++; + } + else + break; + } + while (f); + + return false; +} + +/* Return true if EX excludes F. */ + +bool +excluded_file_name (struct exclude const *ex, char const *f) +{ + struct exclude_segment *seg; + bool invert = false; + char *filename = NULL; + + /* If no patterns are given, the default is to include. */ + if (!ex->head) + return false; + + /* Scan through the segments, reporting the status of the first match. + The segments are in reverse order, so this reports the status of + the last match in the original option list. */ + for (seg = ex->head; ; seg = seg->next) + { + if (seg->type == exclude_hash) + { + if (!filename) + filename = xmalloc (strlen (f) + 1); + if (file_name_matches (seg, f, filename)) + break; + } + else + { + if (file_pattern_matches (seg, f)) + break; + } + + if (! seg->next) + { + /* If patterns are given but none match, the default is the + opposite of the last segment (i.e., the first in the + original option list). For example, in the command + 'grep -r --exclude="a*" --include="*b" pat dir', the + first option is --exclude so any file name matching + neither a* nor *b is included. */ + invert = true; + break; + } + } + + free (filename); + return invert ^ ! (seg->options & EXCLUDE_INCLUDE); +} + +/* Append to EX the exclusion PATTERN with OPTIONS. */ + +void +add_exclude (struct exclude *ex, char const *pattern, int options) +{ + struct exclude_segment *seg; + struct exclude_pattern *pat; + struct patopts *patopts; + + if ((options & (EXCLUDE_REGEX|EXCLUDE_WILDCARDS)) + && fnmatch_pattern_has_wildcards (pattern, options)) + { + if (! (ex->head && ex->head->type == exclude_pattern + && ((ex->head->options & EXCLUDE_INCLUDE) + == (options & EXCLUDE_INCLUDE)))) + new_exclude_segment (ex, exclude_pattern, options); + + seg = ex->head; + + pat = &seg->v.pat; + if (pat->exclude_count == pat->exclude_alloc) + pat->exclude = xpalloc (pat->exclude, &pat->exclude_alloc, 1, -1, + sizeof *pat->exclude); + patopts = &pat->exclude[pat->exclude_count++]; + + patopts->options = options; + if (options & EXCLUDE_REGEX) + { + int rc; + int cflags = REG_NOSUB|REG_EXTENDED| + ((options & FNM_CASEFOLD) ? REG_ICASE : 0); + + if (options & FNM_LEADING_DIR) + { + char *tmp; + idx_t len = strlen (pattern); + + while (len > 0 && ISSLASH (pattern[len-1])) + --len; + + if (len == 0) + rc = 1; + else + { + tmp = ximalloc (len + 7); + memcpy (tmp, pattern, len); + strcpy (tmp + len, "(/.*)?"); + rc = regcomp (&patopts->v.re, tmp, cflags); + free (tmp); + } + } + else + rc = regcomp (&patopts->v.re, pattern, cflags); + + if (rc) + { + pat->exclude_count--; + return; + } + } + else + { + if (options & EXCLUDE_ALLOC) + { + pattern = xstrdup (pattern); + exclude_add_pattern_buffer (ex, (char*) pattern); + } + patopts->v.pattern = pattern; + } + } + else + { + char *str, *p; + int exclude_hash_flags = (EXCLUDE_INCLUDE | EXCLUDE_ANCHORED + | FNM_LEADING_DIR | FNM_CASEFOLD); + if (! (ex->head && ex->head->type == exclude_hash + && ((ex->head->options & exclude_hash_flags) + == (options & exclude_hash_flags)))) + new_exclude_segment (ex, exclude_hash, options); + seg = ex->head; + + str = xstrdup (pattern); + if ((options & (EXCLUDE_WILDCARDS | FNM_NOESCAPE)) == EXCLUDE_WILDCARDS) + unescape_pattern (str); + p = hash_insert (seg->v.table, str); + if (p != str) + free (str); + } +} + +/* Use ADD_FUNC to append to EX the patterns in FILE_NAME, each with + OPTIONS. LINE_END terminates each pattern in the file. If + LINE_END is a space character, ignore trailing spaces and empty + lines in FP. Return -1 (setting errno) on failure, 0 on success. */ + +int +add_exclude_fp (void (*add_func) (struct exclude *, char const *, int, void *), + struct exclude *ex, FILE *fp, int options, + char line_end, + void *data) +{ + char *buf = NULL; + char *p; + char *pattern; + char const *lim; + idx_t buf_alloc = 0; + idx_t buf_count = 0; + int c; + int e = 0; + + while ((c = getc (fp)) != EOF) + { + if (buf_count == buf_alloc) + buf = xpalloc (buf, &buf_alloc, 1, -1, 1); + buf[buf_count++] = c; + } + + if (ferror (fp)) + e = errno; + + buf = xirealloc (buf, buf_count + 1); + buf[buf_count] = line_end; + lim = buf + buf_count + ! (buf_count == 0 || buf[buf_count - 1] == line_end); + + exclude_add_pattern_buffer (ex, buf); + + pattern = buf; + + for (p = buf; p < lim; p++) + if (*p == line_end) + { + char *pattern_end = p; + + if (isspace ((unsigned char) line_end)) + { + for (; ; pattern_end--) + if (pattern_end == pattern) + goto next_pattern; + else if (! isspace ((unsigned char) pattern_end[-1])) + break; + } + + *pattern_end = '\0'; + (*add_func) (ex, pattern, options, data); + + next_pattern: + pattern = p + 1; + } + + errno = e; + return e ? -1 : 0; +} + +static void +call_addfn (struct exclude *ex, char const *pattern, int options, void *data) +{ + void (**addfnptr) (struct exclude *, char const *, int) = data; + (*addfnptr) (ex, pattern, options); +} + +int +add_exclude_file (void (*add_func) (struct exclude *, char const *, int), + struct exclude *ex, char const *file_name, int options, + char line_end) +{ + if (strcmp (file_name, "-") == 0) + return add_exclude_fp (call_addfn, ex, stdin, options, line_end, &add_func); + + FILE *in = fopen (file_name, "re"); + if (!in) + return -1; + int rc = add_exclude_fp (call_addfn, ex, in, options, line_end, &add_func); + int e = errno; + if (fclose (in) != 0) + return -1; + errno = e; + return rc; +} |