diff -u libmagic.orig/apprentice.c libmagic/apprentice.c --- libmagic.orig/apprentice.c 2023-07-17 16:38:35.000000000 +0200 +++ libmagic/apprentice.c 2024-06-09 00:31:40.345830732 +0200 @@ -48,7 +48,9 @@ #ifdef QUICK #include #endif +#ifdef HAVE_DIRENT_H #include +#endif #include #ifdef HAVE_BYTESWAP_H #include @@ -146,10 +148,7 @@ #endif file_private char *mkdbname(struct magic_set *, const char *, int); -file_private struct magic_map *apprentice_buf(struct magic_set *, struct magic *, - size_t); file_private struct magic_map *apprentice_map(struct magic_set *, const char *); -file_private int check_buffer(struct magic_set *, struct magic_map *, const char *); file_private void apprentice_unmap(struct magic_map *); file_private int apprentice_compile(struct magic_set *, struct magic_map *, const char *); @@ -185,6 +184,8 @@ { NULL, 0, NULL } }; +#include "../data_file.c" + #ifdef COMPILE_ONLY int main(int, char *[]); @@ -445,21 +446,13 @@ struct mlist *ml; mlp->map = NULL; - if ((ml = CAST(struct mlist *, malloc(sizeof(*ml)))) == NULL) + if ((ml = CAST(struct mlist *, emalloc(sizeof(*ml)))) == NULL) return -1; ml->map = idx == 0 ? map : NULL; ml->magic = map->magic[idx]; ml->nmagic = map->nmagic[idx]; - if (ml->nmagic) { - ml->magic_rxcomp = CAST(file_regex_t **, - calloc(ml->nmagic, sizeof(*ml->magic_rxcomp))); - if (ml->magic_rxcomp == NULL) { - free(ml); - return -1; - } - } else - ml->magic_rxcomp = NULL; + mlp->prev->next = ml; ml->prev = mlp->prev; ml->next = mlp; @@ -537,13 +530,19 @@ return; for (i = 0; i < MAGIC_SETS; i++) mlist_free(ms->mlist[i]); - free(ms->o.pbuf); - free(ms->o.buf); - free(ms->c.li); + if (ms->o.pbuf) { + efree(ms->o.pbuf); + } + if (ms->o.buf) { + efree(ms->o.buf); + } + if (ms->c.li) { + efree(ms->c.li); + } #ifdef USE_C_LOCALE freelocale(ms->c_lc_ctype); #endif - free(ms); + efree(ms); } file_protected struct magic_set * @@ -552,7 +551,7 @@ struct magic_set *ms; size_t i, len; - if ((ms = CAST(struct magic_set *, calloc(CAST(size_t, 1u), + if ((ms = CAST(struct magic_set *, ecalloc(CAST(size_t, 1u), sizeof(*ms)))) == NULL) return NULL; @@ -565,7 +564,7 @@ ms->o.blen = 0; len = (ms->c.len = 10) * sizeof(*ms->c.li); - if ((ms->c.li = CAST(struct level_info *, malloc(len))) == NULL) + if ((ms->c.li = CAST(struct level_info *, emalloc(len))) == NULL) goto free; ms->event_flags = 0; @@ -589,49 +588,35 @@ #endif return ms; free: - free(ms); + efree(ms); return NULL; } file_private void apprentice_unmap(struct magic_map *map) { - size_t i; - char *p; if (map == NULL) return; - - switch (map->type) { - case MAP_TYPE_USER: - break; - case MAP_TYPE_MALLOC: - p = CAST(char *, map->p); - for (i = 0; i < MAGIC_SETS; i++) { - char *b = RCAST(char *, map->magic[i]); - if (p != NULL && b >= p && b <= p + map->len) - continue; - free(b); + if (map->p != php_magic_database) { + if (map->p == NULL) { + int j; + for (j = 0; j < MAGIC_SETS; j++) { + if (map->magic[j]) { + efree(map->magic[j]); + } + } + } else { + efree(map->p); } - free(p); - break; -#ifdef QUICK - case MAP_TYPE_MMAP: - if (map->p && map->p != MAP_FAILED) - (void)munmap(map->p, map->len); - break; -#endif - default: - fprintf(stderr, "Bad map type %d", map->type); - abort(); } - free(map); + efree(map); } file_private struct mlist * mlist_alloc(void) { struct mlist *mlist; - if ((mlist = CAST(struct mlist *, calloc(1, sizeof(*mlist)))) == NULL) { + if ((mlist = CAST(struct mlist *, ecalloc(1, sizeof(*mlist)))) == NULL) { return NULL; } mlist->next = mlist->prev = mlist; @@ -652,21 +637,9 @@ file_private void mlist_free_one(struct mlist *ml) { - size_t i; - if (ml->map) apprentice_unmap(CAST(struct magic_map *, ml->map)); - - for (i = 0; i < ml->nmagic; ++i) { - if (ml->magic_rxcomp[i]) { - file_regfree(ml->magic_rxcomp[i]); - free(ml->magic_rxcomp[i]); - ml->magic_rxcomp[i] = NULL; - } - } - free(ml->magic_rxcomp); - ml->magic_rxcomp = NULL; - free(ml); + efree(ml); } file_private void @@ -685,50 +658,6 @@ mlist_free_one(mlist); } -#ifndef COMPILE_ONLY -/* void **bufs: an array of compiled magic files */ -file_protected int -buffer_apprentice(struct magic_set *ms, struct magic **bufs, - size_t *sizes, size_t nbufs) -{ - size_t i, j; - struct magic_map *map; - - if (nbufs == 0) - return -1; - - (void)file_reset(ms, 0); - - init_file_tables(); - - for (i = 0; i < MAGIC_SETS; i++) { - mlist_free(ms->mlist[i]); - if ((ms->mlist[i] = mlist_alloc()) == NULL) { - file_oomem(ms, sizeof(*ms->mlist[0])); - goto fail; - } - } - - for (i = 0; i < nbufs; i++) { - map = apprentice_buf(ms, bufs[i], sizes[i]); - if (map == NULL) - goto fail; - - for (j = 0; j < MAGIC_SETS; j++) { - if (add_mlist(ms->mlist[j], map, j) == -1) { - file_oomem(ms, sizeof(*ms->mlist[0])); - goto fail; - } - } - } - - return 0; -fail: - mlist_free_all(ms); - return -1; -} -#endif - /* const char *fn: list of magic files and directories */ file_protected int file_apprentice(struct magic_set *ms, const char *fn, int action) @@ -739,12 +668,28 @@ (void)file_reset(ms, 0); +/* XXX disabling default magic loading so the compiled in data is used */ +#if 0 if ((fn = magic_getpath(fn, action)) == NULL) return -1; +#endif init_file_tables(); - if ((mfn = strdup(fn)) == NULL) { + if (fn == NULL) + fn = getenv("MAGIC"); + if (fn == NULL) { + for (i = 0; i < MAGIC_SETS; i++) { + mlist_free(ms->mlist[i]); + if ((ms->mlist[i] = mlist_alloc()) == NULL) { + file_oomem(ms, sizeof(*ms->mlist[i])); + return -1; + } + } + return apprentice_1(ms, fn, action); + } + + if ((mfn = estrdup(fn)) == NULL) { file_oomem(ms, strlen(fn)); return -1; } @@ -757,7 +702,7 @@ mlist_free(ms->mlist[j]); ms->mlist[j] = NULL; } - free(mfn); + efree(mfn); return -1; } } @@ -774,7 +719,7 @@ fn = p; } - free(mfn); + efree(mfn); if (errs == -1) { for (i = 0; i < MAGIC_SETS; i++) { @@ -1289,7 +1234,7 @@ size_t incr = mset[i].max + ALLOC_INCR; if ((mp = CAST(struct magic_entry *, - realloc(mset[i].me, sizeof(*mp) * incr))) == + erealloc(mset[i].me, sizeof(*mp) * incr))) == NULL) { file_oomem(ms, sizeof(*mp) * incr); return -1; @@ -1312,13 +1257,19 @@ load_1(struct magic_set *ms, int action, const char *fn, int *errs, struct magic_entry_set *mset) { - size_t lineno = 0, llen = 0; + char buffer[BUFSIZ + 1]; char *line = NULL; - ssize_t len; + size_t len; + size_t lineno = 0; struct magic_entry me; - FILE *f = fopen(ms->file = fn, "r"); - if (f == NULL) { + php_stream *stream; + + + ms->file = fn; + stream = php_stream_open_wrapper((char *)fn, "rb", REPORT_ERRORS, NULL); + + if (stream == NULL) { if (errno != ENOENT) file_error(ms, errno, "cannot read magic file `%s'", fn); @@ -1328,8 +1279,7 @@ memset(&me, 0, sizeof(me)); /* read and parse this file */ - for (ms->line = 1; (len = getline(&line, &llen, f)) != -1; - ms->line++) { + for (ms->line = 1; (line = php_stream_get_line(stream, buffer , BUFSIZ, &len)) != NULL; ms->line++) { if (len == 0) /* null line, garbage, etc */ continue; if (line[len - 1] == '\n') { @@ -1388,8 +1338,8 @@ } if (me.mp) (void)addentry(ms, &me, mset); - free(line); - (void)fclose(f); + efree(line); + php_stream_close(stream); } /* @@ -1474,7 +1424,7 @@ } slen = sizeof(**ma) * mentrycount; - if ((*ma = CAST(struct magic *, malloc(slen))) == NULL) { + if ((*ma = CAST(struct magic *, emalloc(slen))) == NULL) { file_oomem(ms, slen); return -1; } @@ -1496,8 +1446,8 @@ if (me == NULL) return; for (i = 0; i < nme; i++) - free(me[i].mp); - free(me); + efree(me[i].mp); + efree(me); } file_private struct magic_map * @@ -1506,18 +1456,19 @@ int errs = 0; uint32_t i, j; size_t files = 0, maxfiles = 0; - char **filearr = NULL, *mfn; - struct stat st; + char **filearr = NULL; + zend_stat_t st = {0}; struct magic_map *map; struct magic_entry_set mset[MAGIC_SETS]; - DIR *dir; - struct dirent *d; + php_stream *dir; + php_stream_dirent d; + memset(mset, 0, sizeof(mset)); ms->flags |= MAGIC_CHECK; /* Enable checks for parsed files */ - if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) + if ((map = CAST(struct magic_map *, ecalloc(1, sizeof(*map)))) == NULL) { file_oomem(ms, sizeof(*map)); return NULL; @@ -1529,52 +1480,50 @@ (void)fprintf(stderr, "%s\n", usg_hdr); /* load directory or file */ - if (stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) { - dir = opendir(fn); + /* FIXME: Read file names and sort them to prevent + non-determinism. See Debian bug #488562. */ + if (php_sys_stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) { + int mflen; + char mfn[MAXPATHLEN]; + + dir = php_stream_opendir((char *)fn, REPORT_ERRORS, NULL); if (!dir) { errs++; goto out; } - while ((d = readdir(dir)) != NULL) { - if (d->d_name[0] == '.') - continue; - if (asprintf(&mfn, "%s/%s", fn, d->d_name) < 0) { + while (php_stream_readdir(dir, &d)) { + if ((mflen = snprintf(mfn, sizeof(mfn), "%s/%s", fn, d.d_name)) < 0) { file_oomem(ms, - strlen(fn) + strlen(d->d_name) + 2); + strlen(fn) + strlen(d.d_name) + 2); errs++; - closedir(dir); + php_stream_closedir(dir); goto out; } - if (stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) { - free(mfn); + if (zend_stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) { continue; } if (files >= maxfiles) { size_t mlen; - char **nfilearr; maxfiles = (maxfiles + 1) * 2; mlen = maxfiles * sizeof(*filearr); - if ((nfilearr = CAST(char **, - realloc(filearr, mlen))) == NULL) { + if ((filearr = CAST(char **, + erealloc(filearr, mlen))) == NULL) { file_oomem(ms, mlen); - free(mfn); - closedir(dir); + php_stream_closedir(dir); errs++; goto out; } - filearr = nfilearr; } - filearr[files++] = mfn; + filearr[files++] = estrndup(mfn, (mflen > sizeof(mfn) - 1)? sizeof(mfn) - 1: mflen); } - closedir(dir); + php_stream_closedir(dir); if (filearr) { qsort(filearr, files, sizeof(*filearr), cmpstrp); for (i = 0; i < files; i++) { load_1(ms, action, filearr[i], &errs, mset); - free(filearr[i]); + efree(filearr[i]); } - free(filearr); - filearr = NULL; + efree(filearr); } } else load_1(ms, action, fn, &errs, mset); @@ -1612,7 +1561,6 @@ } out: - free(filearr); for (j = 0; j < MAGIC_SETS; j++) magic_entry_free(mset[j].me, mset[j].count); @@ -2060,7 +2008,7 @@ if (me->cont_count == me->max_count) { struct magic *nm; size_t cnt = me->max_count + ALLOC_CHUNK; - if ((nm = CAST(struct magic *, realloc(me->mp, + if ((nm = CAST(struct magic *, erealloc(me->mp, sizeof(*nm) * cnt))) == NULL) { file_oomem(ms, sizeof(*nm) * cnt); return -1; @@ -2075,7 +2023,7 @@ static const size_t len = sizeof(*m) * ALLOC_CHUNK; if (me->mp != NULL) return 1; - if ((m = CAST(struct magic *, malloc(len))) == NULL) { + if ((m = CAST(struct magic *, emalloc(len))) == NULL) { file_oomem(ms, len); return -1; } @@ -2301,7 +2249,7 @@ m->mask_op = 0; if (*l == '~') { - if (!IS_STRING(m->type)) + if (!IS_LIBMAGIC_STRING(m->type)) m->mask_op |= FILE_OPINVERSE; else if (ms->flags & MAGIC_CHECK) file_magwarn(ms, "'~' invalid for string types"); @@ -2310,7 +2258,7 @@ m->str_range = 0; m->str_flags = m->type == FILE_PSTRING ? PSTRING_1_LE : 0; if ((op = get_op(*l)) != -1) { - if (IS_STRING(m->type)) { + if (IS_LIBMAGIC_STRING(m->type)) { int r; if (op != FILE_OPDIVIDE) { @@ -2497,7 +2445,7 @@ file_private int parse_extra(struct magic_set *ms, struct magic_entry *me, const char *line, - size_t llen, off_t off, size_t len, const char *name, const char *extra, + size_t llen, zend_off_t off, size_t len, const char *name, const char *extra, int nt) { size_t i; @@ -2860,13 +2808,19 @@ return -1; } if (m->type == FILE_REGEX) { - file_regex_t rx; - int rc = - file_regcomp(ms, &rx, m->value.s, REG_EXTENDED); - if (rc == 0) { - file_regfree(&rx); + zend_string *pattern; + int options = 0; + pcre_cache_entry *pce; + + pattern = convert_libmagic_pattern(m->value.s, strlen(m->value.s), options); + + if ((pce = pcre_get_compiled_regex_cache(pattern)) == NULL) { + zend_string_release(pattern); + return -1; } - return rc ? -1 : 0; + zend_string_release(pattern); + + return 0; } return 0; default: @@ -3245,120 +3199,83 @@ } /* - * handle a buffer containing a compiled file. - */ -file_private struct magic_map * -apprentice_buf(struct magic_set *ms, struct magic *buf, size_t len) -{ - struct magic_map *map; - - if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { - file_oomem(ms, sizeof(*map)); - return NULL; - } - map->len = len; - map->p = buf; - map->type = MAP_TYPE_USER; - if (check_buffer(ms, map, "buffer") != 0) { - apprentice_unmap(map); - return NULL; - } - return map; -} - -/* * handle a compiled file. */ file_private struct magic_map * apprentice_map(struct magic_set *ms, const char *fn) { - int fd; - struct stat st; + uint32_t *ptr; + uint32_t version, entries = 0, nentries; + int needsbyteswap; char *dbname = NULL; struct magic_map *map; - struct magic_map *rv = NULL; + size_t i; + php_stream *stream = NULL; + php_stream_statbuf st; + + - fd = -1; - if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { + if ((map = CAST(struct magic_map *, ecalloc(1, sizeof(*map)))) == NULL) { file_oomem(ms, sizeof(*map)); - goto error; + return NULL; } - map->type = MAP_TYPE_USER; /* unspecified */ + + if (fn == NULL) { + map->p = (void *)&php_magic_database; + goto internal_loaded; + } + +#ifdef PHP_WIN32 + /* Don't bother on windows with php_stream_open_wrapper, + return to give apprentice_load() a chance. */ + if (php_stream_stat_path_ex((char *)fn, 0, &st, NULL) == SUCCESS) { + if (st.sb.st_mode & S_IFDIR) { + goto error; + } + } +#endif dbname = mkdbname(ms, fn, 0); if (dbname == NULL) goto error; - if ((fd = open(dbname, O_RDONLY|O_BINARY)) == -1) + stream = php_stream_open_wrapper((char *)fn, "rb", REPORT_ERRORS, NULL); + + if (!stream) { goto error; + } - if (fstat(fd, &st) == -1) { +#ifndef PHP_WIN32 + if (php_stream_stat(stream, &st) < 0) { file_error(ms, errno, "cannot stat `%s'", dbname); goto error; } - if (st.st_size < 8 || st.st_size > maxoff_t()) { +#endif + if (st.sb.st_size < 8 || st.sb.st_size > maxoff_t()) { file_error(ms, 0, "file `%s' is too %s", dbname, - st.st_size < 8 ? "small" : "large"); + st.sb.st_size < 8 ? "small" : "large"); goto error; } - map->len = CAST(size_t, st.st_size); -#ifdef QUICK - map->type = MAP_TYPE_MMAP; - if ((map->p = mmap(0, CAST(size_t, st.st_size), PROT_READ|PROT_WRITE, - MAP_PRIVATE|MAP_FILE, fd, CAST(off_t, 0))) == MAP_FAILED) { - file_error(ms, errno, "cannot map `%s'", dbname); - goto error; - } -#else map->type = MAP_TYPE_MALLOC; - if ((map->p = CAST(void *, malloc(map->len))) == NULL) { - file_oomem(ms, map->len); - goto error; - } - if (read(fd, map->p, map->len) != (ssize_t)map->len) { - file_badread(ms); - goto error; - } -#endif - (void)close(fd); - fd = -1; + map->len = CAST(size_t, st.sb.st_size); + map->p = CAST(void *, emalloc(map->len)); - if (check_buffer(ms, map, dbname) != 0) { - goto error; - } -#ifdef QUICK - if (mprotect(map->p, CAST(size_t, st.st_size), PROT_READ) == -1) { - file_error(ms, errno, "cannot mprotect `%s'", dbname); + if (php_stream_read(stream, map->p, (size_t)st.sb.st_size) != (size_t)st.sb.st_size) { + file_badread(ms); goto error; } -#endif - - free(dbname); - return map; - -error: - if (fd != -1) - (void)close(fd); - apprentice_unmap(map); - free(dbname); - return rv; -} -file_private int -check_buffer(struct magic_set *ms, struct magic_map *map, const char *dbname) -{ - uint32_t *ptr; - uint32_t entries, nentries; - uint32_t version; - int i, needsbyteswap; + php_stream_close(stream); + stream = NULL; - ptr = CAST(uint32_t *, map->p); +internal_loaded: + ptr = (uint32_t *)(void *)map->p; if (*ptr != MAGICNO) { if (swap4(*ptr) != MAGICNO) { file_error(ms, 0, "bad magic in `%s'", dbname); - return -1; + goto error; } needsbyteswap = 1; } else @@ -3368,17 +3285,29 @@ else version = ptr[1]; if (version != VERSIONNO) { - file_error(ms, 0, "File %s supports only version %d magic " - "files. `%s' is version %d", VERSION, + file_error(ms, 0, "File %d supports only version %d magic " + "files. `%s' is version %d", MAGIC_VERSION, VERSIONNO, dbname, version); - return -1; + goto error; } - entries = CAST(uint32_t, map->len / sizeof(struct magic)); - if ((entries * sizeof(struct magic)) != map->len) { - file_error(ms, 0, "Size of `%s' %" SIZE_T_FORMAT "u is not " - "a multiple of %" SIZE_T_FORMAT "u", - dbname, map->len, sizeof(struct magic)); - return -1; + + /* php_magic_database is a const, performing writes will segfault. This is for big-endian + machines only, PPC and Sparc specifically. Consider static variable or MINIT in + future. */ + if (needsbyteswap && fn == NULL) { + map->p = emalloc(sizeof(php_magic_database)); + map->p = memcpy(map->p, php_magic_database, sizeof(php_magic_database)); + } + + if (NULL != fn) { + nentries = (uint32_t)(st.sb.st_size / sizeof(struct magic)); + entries = (uint32_t)(st.sb.st_size / sizeof(struct magic)); + if ((zend_off_t)(entries * sizeof(struct magic)) != st.sb.st_size) { + file_error(ms, 0, "Size of `%s' %llu is not a multiple of %zu", + dbname, (unsigned long long)st.sb.st_size, + sizeof(struct magic)); + goto error; + } } map->magic[0] = CAST(struct magic *, map->p) + 1; nentries = 0; @@ -3391,15 +3320,29 @@ map->magic[i + 1] = map->magic[i] + map->nmagic[i]; nentries += map->nmagic[i]; } - if (entries != nentries + 1) { + if (NULL != fn && entries != nentries + 1) { file_error(ms, 0, "Inconsistent entries in `%s' %u != %u", dbname, entries, nentries + 1); - return -1; + goto error; } if (needsbyteswap) for (i = 0; i < MAGIC_SETS; i++) byteswap(map->magic[i], map->nmagic[i]); - return 0; + + if (dbname) { + efree(dbname); + } + return map; + +error: + if (stream) { + php_stream_close(stream); + } + apprentice_unmap(map); + if (dbname) { + efree(dbname); + } + return NULL; } /* @@ -3410,7 +3353,7 @@ { static const size_t nm = sizeof(*map->nmagic) * MAGIC_SETS; static const size_t m = sizeof(**map->magic); - int fd = -1; + php_stream *stream; size_t len; char *dbname; int rv = -1; @@ -3425,8 +3368,10 @@ if (dbname == NULL) goto out; - if ((fd = open(dbname, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY, 0644)) == -1) - { + /* wb+ == O_WRONLY|O_CREAT|O_TRUNC|O_BINARY */ + stream = php_stream_open_wrapper((char *)fn, "wb+", REPORT_ERRORS, NULL); + + if (!stream) { file_error(ms, errno, "cannot open `%s'", dbname); goto out; } @@ -3435,26 +3380,25 @@ hdr.h[1] = VERSIONNO; memcpy(hdr.h + 2, map->nmagic, nm); - if (write(fd, &hdr, sizeof(hdr)) != CAST(ssize_t, sizeof(hdr))) { + if (php_stream_write(stream,(const char *)&hdr, sizeof(hdr)) != (ssize_t)sizeof(hdr)) { file_error(ms, errno, "error writing `%s'", dbname); - goto out2; + goto out; } for (i = 0; i < MAGIC_SETS; i++) { len = m * map->nmagic[i]; - if (write(fd, map->magic[i], len) != CAST(ssize_t, len)) { + if (php_stream_write(stream, (const char *)map->magic[i], len) != (ssize_t)len) { file_error(ms, errno, "error writing `%s'", dbname); - goto out2; + goto out; } } rv = 0; -out2: - if (fd != -1) - (void)close(fd); + if (stream) { + php_stream_close(stream); + } out: - apprentice_unmap(map); - free(dbname); + efree(dbname); return rv; } @@ -3488,17 +3432,18 @@ q++; /* Compatibility with old code that looked in .mime */ if (ms->flags & MAGIC_MIME) { - if (asprintf(&buf, "%.*s.mime%s", CAST(int, q - fn), fn, ext) - < 0) - return NULL; - if (access(buf, R_OK) != -1) { + spprintf(&buf, MAXPATHLEN, "%.*s.mime%s", CAST(int, q - fn), fn, ext); +#ifdef PHP_WIN32 + if (VCWD_ACCESS(buf, R_OK) == 0) { +#else + if (VCWD_ACCESS(buf, R_OK) != -1) { +#endif ms->flags &= MAGIC_MIME_TYPE; return buf; } - free(buf); + efree(buf); } - if (asprintf(&buf, "%.*s%s", CAST(int, q - fn), fn, ext) < 0) - return NULL; + spprintf(&buf, MAXPATHLEN, "%.*s%s", CAST(int, q - fn), fn, ext); /* Compatibility with old code that looked in .mime */ if (strstr(fn, ".mime") != NULL) @@ -3620,7 +3565,7 @@ m->offset = swap4(CAST(uint32_t, m->offset)); m->in_offset = swap4(CAST(uint32_t, m->in_offset)); m->lineno = swap4(CAST(uint32_t, m->lineno)); - if (IS_STRING(m->type)) { + if (IS_LIBMAGIC_STRING(m->type)) { m->str_range = swap4(m->str_range); m->str_flags = swap4(m->str_flags); } @@ -3717,7 +3662,6 @@ continue; if (strcmp(ma[i].value.s, name) == 0) { v->magic = &ma[i]; - v->magic_rxcomp = &(ml->magic_rxcomp[i]); for (j = i + 1; j < ml->nmagic; j++) if (ma[j].cont_level == 0) break; diff -u libmagic.orig/ascmagic.c libmagic/ascmagic.c --- libmagic.orig/ascmagic.c 2023-05-30 22:17:50.000000000 +0200 +++ libmagic/ascmagic.c 2024-06-09 00:31:40.345830732 +0200 @@ -96,7 +96,7 @@ rv = file_ascmagic_with_encoding(ms, &bb, ubuf, ulen, code, type, text); - free(ubuf); + efree(ubuf); return rv; } @@ -143,7 +143,7 @@ /* malloc size is a conservative overestimate; could be improved, or at least realloced after conversion. */ mlen = ulen * 6; - if ((utf8_buf = CAST(unsigned char *, malloc(mlen))) == NULL) { + if ((utf8_buf = CAST(unsigned char *, emalloc(mlen))) == NULL) { file_oomem(ms, mlen); goto done; } @@ -267,8 +267,8 @@ goto done; if (has_long_lines) - if (file_printf(ms, ", with very long lines (%" - SIZE_T_FORMAT "u)", has_long_lines) == -1) + if (file_printf(ms, ", with very long lines (%zu)", + has_long_lines) == -1) goto done; /* @@ -324,7 +324,8 @@ } rv = 1; done: - free(utf8_buf); + if (utf8_buf) + efree(utf8_buf); return rv; } diff -u libmagic.orig/buffer.c libmagic/buffer.c --- libmagic.orig/buffer.c 2023-07-02 14:48:39.000000000 +0200 +++ libmagic/buffer.c 2024-06-09 00:31:40.345830732 +0200 @@ -31,19 +31,21 @@ #endif /* lint */ #include "magic.h" +#ifdef HAVE_UNISTD_H #include +#endif #include #include #include void -buffer_init(struct buffer *b, int fd, const struct stat *st, const void *data, +buffer_init(struct buffer *b, int fd, const zend_stat_t *st, const void *data, size_t len) { b->fd = fd; if (st) memcpy(&b->st, st, sizeof(b->st)); - else if (b->fd == -1 || fstat(b->fd, &b->st) == -1) + else if (b->fd == -1 || zend_fstat(b->fd, &b->st) == -1) memset(&b->st, 0, sizeof(b->st)); b->fbuf = data; b->flen = len; @@ -55,7 +57,7 @@ void buffer_fini(struct buffer *b) { - free(b->ebuf); + efree(b->ebuf); b->ebuf = NULL; b->elen = 0; } @@ -74,16 +76,18 @@ b->elen = CAST(size_t, b->st.st_size) < b->flen ? CAST(size_t, b->st.st_size) : b->flen; if (b->elen == 0) { - free(b->ebuf); + efree(b->ebuf); b->ebuf = NULL; return 0; } - if ((b->ebuf = malloc(b->elen)) == NULL) + if ((b->ebuf = emalloc(b->elen)) == NULL) goto out; b->eoff = b->st.st_size - b->elen; - if (pread(b->fd, b->ebuf, b->elen, b->eoff) == -1) { - free(b->ebuf); + if (FINFO_LSEEK_FUNC(b->fd, b->eoff, SEEK_SET) == (zend_off_t)-1 || + FINFO_READ_FUNC(b->fd, b->ebuf, b->elen) != (ssize_t)b->elen) + { + efree(b->ebuf); b->ebuf = NULL; goto out; } diff -u libmagic.orig/cdf.c libmagic/cdf.c --- libmagic.orig/cdf.c 2022-09-24 22:56:49.000000000 +0200 +++ libmagic/cdf.c 2024-06-09 00:31:40.345830732 +0200 @@ -43,7 +43,9 @@ #include #endif #include +#ifdef HAVE_UNISTD_H #include +#endif #include #include #include @@ -91,44 +93,9 @@ CDF_TOLE8(CAST(uint64_t, x)))) #define CDF_GETUINT32(x, y) cdf_getuint32(x, y) -#define CDF_MALLOC(n) cdf_malloc(__FILE__, __LINE__, (n)) -#define CDF_REALLOC(p, n) cdf_realloc(__FILE__, __LINE__, (p), (n)) -#define CDF_CALLOC(n, u) cdf_calloc(__FILE__, __LINE__, (n), (u)) - - -/*ARGSUSED*/ -static void * -cdf_malloc(const char *file __attribute__((__unused__)), - size_t line __attribute__((__unused__)), size_t n) -{ - DPRINTF(("%s,%" SIZE_T_FORMAT "u: %s %" SIZE_T_FORMAT "u\n", - file, line, __func__, n)); - if (n == 0) - n++; - return malloc(n); -} - -/*ARGSUSED*/ -static void * -cdf_realloc(const char *file __attribute__((__unused__)), - size_t line __attribute__((__unused__)), void *p, size_t n) -{ - DPRINTF(("%s,%" SIZE_T_FORMAT "u: %s %" SIZE_T_FORMAT "u\n", - file, line, __func__, n)); - return realloc(p, n); -} - -/*ARGSUSED*/ -static void * -cdf_calloc(const char *file __attribute__((__unused__)), - size_t line __attribute__((__unused__)), size_t n, size_t u) -{ - DPRINTF(("%s,%" SIZE_T_FORMAT "u: %s %" SIZE_T_FORMAT "u %" - SIZE_T_FORMAT "u\n", file, line, __func__, n, u)); - if (n == 0) - n++; - return calloc(n, u); -} +#define CDF_MALLOC(n) emalloc(n) +#define CDF_REALLOC(p, n) erealloc(p, n) +#define CDF_CALLOC(n, u) ecalloc(n, u) #if defined(HAVE_BYTESWAP_H) # define _cdf_tole2(x) bswap_16(x) @@ -334,7 +301,7 @@ scn->sst_len = 0; scn->sst_dirlen = 0; scn->sst_ss = 0; - free(scn->sst_tab); + efree(scn->sst_tab); scn->sst_tab = NULL; return -1; } @@ -342,9 +309,11 @@ static size_t cdf_check_stream(const cdf_stream_t *sst, const cdf_header_t *h) { +#ifndef NDEBUG size_t ss = sst->sst_dirlen < h->h_min_size_standard_stream ? CDF_SHORT_SEC_SIZE(h) : CDF_SEC_SIZE(h); assert(ss == sst->sst_ss); +#endif return sst->sst_ss; } @@ -367,11 +336,11 @@ } static ssize_t -cdf_read(const cdf_info_t *info, off_t off, void *buf, size_t len) +cdf_read(const cdf_info_t *info, zend_off_t off, void *buf, size_t len) { size_t siz = CAST(size_t, off + len); - if (CAST(off_t, off + len) != CAST(off_t, siz)) + if (CAST(zend_off_t, off + len) != CAST(zend_off_t, siz)) goto out; if (info->i_buf != NULL && info->i_len >= siz) { @@ -382,7 +351,10 @@ if (info->i_fd == -1) goto out; - if (pread(info->i_fd, buf, len, off) != CAST(ssize_t, len)) + if (FINFO_LSEEK_FUNC(info->i_fd, off, SEEK_SET) == (zend_off_t)-1) + return -1; + + if (FINFO_READ_FUNC(info->i_fd, buf, len) != (ssize_t)len) return -1; return CAST(ssize_t, len); @@ -397,7 +369,7 @@ char buf[512]; (void)memcpy(cdf_bo.s, "\01\02\03\04", 4); - if (cdf_read(info, CAST(off_t, 0), buf, sizeof(buf)) == -1) + if (cdf_read(info, CAST(zend_off_t, 0), buf, sizeof(buf)) == -1) return -1; cdf_unpack_header(h, buf); cdf_swap_header(h); @@ -544,14 +516,14 @@ } out: sat->sat_len = i; - free(msa); + efree(msa); return 0; out3: errno = EFTYPE; out2: - free(msa); + efree(msa); out1: - free(sat->sat_tab); + efree(sat->sat_tab); return -1; } @@ -719,7 +691,7 @@ return -1; if ((buf = CAST(char *, CDF_MALLOC(ss))) == NULL) { - free(dir->dir_tab); + efree(dir->dir_tab); return -1; } @@ -742,11 +714,11 @@ if (NEED_SWAP) for (i = 0; i < dir->dir_len; i++) cdf_swap_dir(&dir->dir_tab[i]); - free(buf); + efree(buf); return 0; out: - free(dir->dir_tab); - free(buf); + efree(dir->dir_tab); + efree(buf); errno = EFTYPE; return -1; } @@ -791,7 +763,7 @@ out: errno = EFTYPE; out1: - free(ssat->sat_tab); + efree(ssat->sat_tab); return -1; } @@ -953,7 +925,7 @@ *maxcount = newcount; return inp; out: - free(*info); + efree(*info); *maxcount = 0; *info = NULL; return NULL; @@ -1136,7 +1108,7 @@ } return 0; out: - free(*info); + efree(*info); *info = NULL; *count = 0; *maxcount = 0; @@ -1428,7 +1400,7 @@ cdf_directory_t *d; char name[__arraycount(d->d_name)]; cdf_stream_t scn; - struct timespec ts; + struct timeval ts; static const char *types[] = { "empty", "user storage", "user stream", "lockbytes", "property", "root storage" }; @@ -1470,7 +1442,7 @@ break; } cdf_dump_stream(&scn); - free(scn.sst_tab); + efree(scn.sst_tab); break; default: break; @@ -1483,7 +1455,7 @@ cdf_dump_property_info(const cdf_property_info_t *info, size_t count) { cdf_timestamp_t tp; - struct timespec ts; + struct timeval ts; char buf[64]; size_t i, j; @@ -1568,7 +1540,7 @@ (void)fprintf(stderr, "Class %s\n", buf); (void)fprintf(stderr, "Count %d\n", ssi.si_count); cdf_dump_property_info(info, count); - free(info); + efree(info); } @@ -1589,7 +1561,7 @@ cdf_u16tos8(sbuf, ce[i].ce_namlen, ce[i].ce_name), cdf_ctime(&ts.tv_sec, tbuf)); } - free(cat); + efree(cat); } #endif diff -u libmagic.orig/cdf.h libmagic/cdf.h --- libmagic.orig/cdf.h 2022-09-24 22:56:49.000000000 +0200 +++ libmagic/cdf.h 2024-06-04 15:10:40.600783222 +0200 @@ -37,8 +37,6 @@ #ifdef WIN32 #include -#define timespec timeval -#define tv_nsec tv_usec #endif #ifdef __DJGPP__ #define timespec timeval diff -u libmagic.orig/compress.c libmagic/compress.c --- libmagic.orig/compress.c 2023-05-21 17:59:58.000000000 +0200 +++ libmagic/compress.c 2024-06-09 00:31:40.346830746 +0200 @@ -63,13 +63,14 @@ #if defined(HAVE_SYS_TIME_H) #include #endif - -#if defined(HAVE_ZLIB_H) && defined(ZLIBSUPPORT) +#if defined(HAVE_ZLIB_H) && defined(PHP_FILEINFO_UNCOMPRESS) #define BUILTIN_DECOMPRESS #include #endif -#if defined(HAVE_BZLIB_H) && defined(BZLIBSUPPORT) +#undef FIONREAD + +#if defined(PHP_FILEINFO_UNCOMPRESS) #define BUILTIN_BZLIB #include #endif @@ -132,6 +133,8 @@ } #endif +#ifdef PHP_FILEINFO_UNCOMPRESS + static int lzmacmp(const unsigned char *buf) { @@ -307,7 +310,7 @@ } nsz = nbytes; - free(newbuf); + efree(newbuf); urv = uncompressbuf(fd, ms->bytes_max, i, (ms->flags & MAGIC_NO_COMPRESS_FORK), buf, &newbuf, &nsz); DPRINTF("uncompressbuf = %d, %s, %" SIZE_T_FORMAT "u\n", urv, @@ -319,7 +322,7 @@ if (urv == ERRDATA) prv = format_decompression_error(ms, i, newbuf); else - prv = file_buffer(ms, -1, NULL, name, newbuf, + prv = file_buffer(ms, NULL, NULL, name, newbuf, nsz); if (prv == -1) goto error; @@ -337,7 +340,7 @@ * XXX: If file_buffer fails here, we overwrite * the compressed text. FIXME. */ - if (file_buffer(ms, -1, NULL, NULL, buf, nbytes) == -1) + if (file_buffer(ms, NULL, NULL, NULL, buf, nbytes) == -1) { if (file_pop_buffer(ms, pb) != NULL) abort(); @@ -345,10 +348,10 @@ } if ((rbuf = file_pop_buffer(ms, pb)) != NULL) { if (file_printf(ms, "%s", rbuf) == -1) { - free(rbuf); + efree(rbuf); goto error; } - free(rbuf); + efree(rbuf); } if (!mime && file_printf(ms, ")") == -1) goto error; @@ -369,7 +372,8 @@ if (sa_saved && sig_act.sa_handler != SIG_IGN) (void)sigaction(SIGPIPE, &sig_act, NULL); - free(newbuf); + if (newbuf) + efree(newbuf); ms->flags |= MAGIC_COMPRESS; DPRINTF("Zmagic returns %d\n", rv); return rv; @@ -452,7 +456,7 @@ nocheck: do - switch ((rv = read(fd, buf, n))) { + switch ((rv = FINFO_READ_FUNC(fd, buf, n))) { case -1: if (errno == EINTR) continue; @@ -545,13 +549,13 @@ return -1; } (void)close(tfd); - if (lseek(fd, CAST(off_t, 0), SEEK_SET) == CAST(off_t, -1)) { + if (FINFO_LSEEK_FUNC(fd, (zend_off_t)0, SEEK_SET) == (zend_off_t)-1) { file_badseek(ms); return -1; } return fd; } -#if HAVE_FORK +#ifdef PHP_FILEINFO_UNCOMPRESS #ifdef BUILTIN_DECOMPRESS #define FHCRC (1 << 1) @@ -1062,7 +1066,7 @@ int (*decompress)(const unsigned char *, unsigned char **, size_t, size_t *, int) = getdecompressor(method); - *newch = CAST(unsigned char *, malloc(bytes_max + 1)); + *newch = CAST(unsigned char *, emalloc(bytes_max + 1)); if (*newch == NULL) return makeerror(newch, n, "No buffer, %s", strerror(errno)); @@ -1225,3 +1229,4 @@ return rv; } #endif +#endif diff -u libmagic.orig/der.c libmagic/der.c --- libmagic.orig/der.c 2022-09-24 22:56:49.000000000 +0200 +++ libmagic/der.c 2024-06-09 00:31:40.346830746 +0200 @@ -54,7 +54,9 @@ #include "magic.h" #include "der.h" #else +#ifdef HAVE_SYS_MMAN_H #include +#endif #include #include #endif diff -u libmagic.orig/encoding.c libmagic/encoding.c --- libmagic.orig/encoding.c 2022-12-26 18:31:56.000000000 +0100 +++ libmagic/encoding.c 2024-06-09 00:31:40.346830746 +0200 @@ -97,7 +97,7 @@ nbytes = ms->encoding_max; mlen = (nbytes + 1) * sizeof((*ubuf)[0]); - *ubuf = CAST(file_unichar_t *, calloc(CAST(size_t, 1), mlen)); + *ubuf = CAST(file_unichar_t *, ecalloc(CAST(size_t, 1), mlen)); if (*ubuf == NULL) { file_oomem(ms, mlen); goto done; @@ -150,7 +150,7 @@ unsigned char *nbuf; mlen = (nbytes + 1) * sizeof(nbuf[0]); - if ((nbuf = CAST(unsigned char *, malloc(mlen))) == NULL) { + if ((nbuf = CAST(unsigned char *, emalloc(mlen))) == NULL) { file_oomem(ms, mlen); goto done; } @@ -170,12 +170,12 @@ rv = 0; *type = "binary"; } - free(nbuf); + efree(nbuf); } done: if (ubuf == &udefbuf) - free(udefbuf); + efree(udefbuf); return rv; } diff -u libmagic.orig/file.h libmagic/file.h --- libmagic.orig/file.h 2023-07-27 21:40:22.000000000 +0200 +++ libmagic/file.h 2024-06-09 00:31:40.346830746 +0200 @@ -27,15 +27,13 @@ */ /* * file.h - definitions for file(1) program - * @(#)$File: file.h,v 1.247 2023/07/27 19:40:22 christos Exp $ + * @(#)$File: file.h,v 1.248 2023/07/28 14:38:25 christos Exp $ */ #ifndef __file_h__ #define __file_h__ -#ifdef HAVE_CONFIG_H -#include -#endif +#include "config.h" #ifdef HAVE_STDINT_H #include @@ -79,8 +77,7 @@ #include /* Include that here, to make sure __P gets defined */ #include #include /* For open and flags */ -#include -#include + #include #ifndef WIN32 #include @@ -88,10 +85,6 @@ /* Do this here and now, because struct stat gets re-defined on solaris */ #include #include -#include -#if defined(HAVE_XLOCALE_H) -#include -#endif #define ENABLE_CONDITIONALS @@ -159,9 +152,11 @@ /* * Dec 31, 23:59:59 9999 * we need to make sure that we don't exceed 9999 because some libc - * implementations like muslc crash otherwise + * implementations like muslc crash otherwise. If you are unlucky + * to be running on a system with a 32 bit time_t, then it is even less. */ -#define MAX_CTIME CAST(time_t, 0x3afff487cfULL) +#define MAX_CTIME \ + CAST(time_t, sizeof(time_t) > 4 ? 0x3afff487cfULL : 0x7fffffffULL) #define FILE_BADSIZE CAST(size_t, ~0ul) #define MAXDESC 64 /* max len of text description/MIME type */ @@ -179,14 +174,12 @@ #define FILE_COMPILE 2 #define FILE_LIST 3 -typedef regex_t file_regex_t; - struct buffer { int fd; - struct stat st; + zend_stat_t st; const void *fbuf; size_t flen; - off_t eoff; + zend_off_t eoff; void *ebuf; size_t elen; }; @@ -289,7 +282,7 @@ #define FILE_OCTAL 59 #define FILE_NAMES_SIZE 60 /* size of array to contain all names */ -#define IS_STRING(t) \ +#define IS_LIBMAGIC_STRING(t) \ ((t) == FILE_STRING || \ (t) == FILE_PSTRING || \ (t) == FILE_BESTRING16 || \ @@ -420,7 +413,6 @@ /* list of magic entries */ struct mlist { struct magic *magic; /* array of magic entries */ - file_regex_t **magic_rxcomp; /* array of compiled regexps */ size_t nmagic; /* number of entries in array */ void *map; /* internal resources used by entry */ struct mlist *next, *prev; @@ -525,11 +517,9 @@ file_protected const char *file_fmtnum(char *, size_t, const char *, int); file_protected struct magic_set *file_ms_alloc(int); file_protected void file_ms_free(struct magic_set *); -file_protected int file_default(struct magic_set *, size_t); -file_protected int file_buffer(struct magic_set *, int, struct stat *, - const char *, const void *, size_t); -file_protected int file_fsmagic(struct magic_set *, const char *, - struct stat *); +file_protected int file_buffer(struct magic_set *, php_stream *, zend_stat_t *, const char *, const void *, + size_t); +file_protected int file_fsmagic(struct magic_set *, const char *, zend_stat_t *); file_protected int file_pipe2file(struct magic_set *, int, const void *, size_t); file_protected int file_vprintf(struct magic_set *, const char *, va_list) @@ -546,7 +536,7 @@ file_protected int file_reset(struct magic_set *, int); file_protected int file_tryelf(struct magic_set *, const struct buffer *); file_protected int file_trycdf(struct magic_set *, const struct buffer *); -#if HAVE_FORK +#ifdef PHP_FILEINFO_UNCOMPRESS file_protected int file_zmagic(struct magic_set *, const struct buffer *, const char *); #endif @@ -605,19 +595,13 @@ file_protected int file_clear_closexec(int); file_protected char *file_strtrim(char *); -file_protected void buffer_init(struct buffer *, int, const struct stat *, +file_protected void buffer_init(struct buffer *, int, const zend_stat_t *, const void *, size_t); file_protected void buffer_fini(struct buffer *); file_protected int buffer_fill(const struct buffer *); -file_protected int file_regcomp(struct magic_set *, file_regex_t *, - const char *, int); -file_protected int file_regexec(struct magic_set *, file_regex_t *, - const char *, size_t, regmatch_t *, int); -file_protected void file_regfree(file_regex_t *); - typedef struct { char *buf; size_t blen; @@ -632,23 +616,10 @@ extern const size_t file_nnames; #endif -#ifndef HAVE_PREAD -ssize_t pread(int, void *, size_t, off_t); -#endif -#ifndef HAVE_VASPRINTF -int vasprintf(char **, const char *, va_list); -#endif -#ifndef HAVE_ASPRINTF -int asprintf(char **, const char *, ...); -#endif -#ifndef HAVE_DPRINTF -int dprintf(int, const char *, ...); -#endif - -#ifndef HAVE_STRLCPY +#ifndef strlcpy size_t strlcpy(char *, const char *, size_t); #endif -#ifndef HAVE_STRLCAT +#ifndef strlcat size_t strlcat(char *, const char *, size_t); #endif #ifndef HAVE_STRCASESTR @@ -664,39 +635,6 @@ #ifndef HAVE_ASCTIME_R char *asctime_r(const struct tm *, char *); #endif -#ifndef HAVE_GMTIME_R -struct tm *gmtime_r(const time_t *, struct tm *); -#endif -#ifndef HAVE_LOCALTIME_R -struct tm *localtime_r(const time_t *, struct tm *); -#endif -#ifndef HAVE_FMTCHECK -const char *fmtcheck(const char *, const char *) - __attribute__((__format_arg__(2))); -#endif - -#ifdef HAVE_LIBSECCOMP -// basic filter -// this mode should not interfere with normal operations -// only some dangerous syscalls are blacklisted -int enable_sandbox_basic(void); - -// enhanced filter -// this mode allows only the necessary syscalls used during normal operation -// extensive testing required !!! -int enable_sandbox_full(void); -#endif - -file_protected const char *file_getprogname(void); -file_protected void file_setprogname(const char *); -file_protected void file_err(int, const char *, ...) - __attribute__((__format__(__printf__, 2, 3), __noreturn__)); -file_protected void file_errx(int, const char *, ...) - __attribute__((__format__(__printf__, 2, 3), __noreturn__)); -file_protected void file_warn(const char *, ...) - __attribute__((__format__(__printf__, 1, 2))); -file_protected void file_warnx(const char *, ...) - __attribute__((__format__(__printf__, 1, 2))); #if defined(HAVE_MMAP) && defined(HAVE_SYS_MMAN_H) && !defined(QUICK) #define QUICK diff -u libmagic.orig/fsmagic.c libmagic/fsmagic.c --- libmagic.orig/fsmagic.c 2023-07-27 21:33:24.000000000 +0200 +++ libmagic/fsmagic.c 2024-06-09 00:31:40.346830746 +0200 @@ -66,26 +66,10 @@ # define minor(dev) ((dev) & 0xff) #endif #undef HAVE_MAJOR -#ifdef S_IFLNK -file_private int -bad_link(struct magic_set *ms, int err, char *buf) -{ - int mime = ms->flags & MAGIC_MIME; - if ((mime & MAGIC_MIME_TYPE) && - file_printf(ms, "inode/symlink") - == -1) - return -1; - else if (!mime) { - if (ms->flags & MAGIC_ERROR) { - file_error(ms, err, - "broken symbolic link to %s", buf); - return -1; - } - if (file_printf(ms, "broken symbolic link to %s", buf) == -1) - return -1; - } - return 1; -} + +#ifdef PHP_WIN32 + +# undef S_IFIFO #endif file_private int handle_mime(struct magic_set *ms, int mime, const char *str) @@ -103,60 +87,17 @@ } file_protected int -file_fsmagic(struct magic_set *ms, const char *fn, struct stat *sb) +file_fsmagic(struct magic_set *ms, const char *fn, zend_stat_t *sb) { int ret, did = 0; int mime = ms->flags & MAGIC_MIME; int silent = ms->flags & (MAGIC_APPLE|MAGIC_EXTENSION); -#ifdef S_IFLNK - char buf[BUFSIZ+4]; - ssize_t nch; - struct stat tstatbuf; -#endif if (fn == NULL) return 0; #define COMMA (did++ ? ", " : "") - /* - * Fstat is cheaper but fails for files you don't have read perms on. - * On 4.2BSD and similar systems, use lstat() to identify symlinks. - */ -#ifdef S_IFLNK - if ((ms->flags & MAGIC_SYMLINK) == 0) - ret = lstat(fn, sb); - else -#endif - ret = stat(fn, sb); /* don't merge into if; see "ret =" above */ - -#ifdef WIN32 - { - HANDLE hFile = CreateFile((LPCSTR)fn, 0, FILE_SHARE_DELETE | - FILE_SHARE_READ | FILE_SHARE_WRITE, NULL, OPEN_EXISTING, 0, - NULL); - if (hFile != INVALID_HANDLE_VALUE) { - /* - * Stat failed, but we can still open it - assume it's - * a block device, if nothing else. - */ - if (ret) { - sb->st_mode = S_IFBLK; - ret = 0; - } - switch (GetFileType(hFile)) { - case FILE_TYPE_CHAR: - sb->st_mode |= S_IFCHR; - sb->st_mode &= ~S_IFREG; - break; - case FILE_TYPE_PIPE: - sb->st_mode |= S_IFIFO; - sb->st_mode &= ~S_IFREG; - break; - } - CloseHandle(hFile); - } - } -#endif + ret = php_sys_stat(fn, sb); if (ret) { if (ms->flags & MAGIC_ERROR) { @@ -189,32 +130,24 @@ } switch (sb->st_mode & S_IFMT) { - case S_IFDIR: - if (mime) { - if (handle_mime(ms, mime, "directory") == -1) - return -1; - } else if (silent) { - } else if (file_printf(ms, "%sdirectory", COMMA) == -1) - return -1; - break; -#ifdef S_IFCHR - case S_IFCHR: - /* - * If -s has been specified, treat character special files - * like ordinary files. Otherwise, just report that they - * are block special files and go on to the next file. - */ - if ((ms->flags & MAGIC_DEVICES) != 0) { - ret = 0; - break; - } - if (mime) { - if (handle_mime(ms, mime, "chardevice") == -1) - return -1; - } else if (silent) { - } else { -#ifdef HAVE_STRUCT_STAT_ST_RDEV -# ifdef dv_unit +#ifndef PHP_WIN32 +# ifdef S_IFCHR + case S_IFCHR: + /* + * If -s has been specified, treat character special files + * like ordinary files. Otherwise, just report that they + * are block special files and go on to the next file. + */ + if ((ms->flags & MAGIC_DEVICES) != 0) { + ret = 0; + break; + } + if (mime) { + if (handle_mime(ms, mime, "chardevice") == -1) + return -1; + } else { +# ifdef HAVE_STAT_ST_RDEV +# ifdef dv_unit if (file_printf(ms, "%scharacter special (%d/%d/%d)", COMMA, major(sb->st_rdev), dv_unit(sb->st_rdev), dv_subunit(sb->st_rdev)) == -1) @@ -229,45 +162,11 @@ if (file_printf(ms, "%scharacter special", COMMA) == -1) return -1; #endif - } - break; -#endif -#ifdef S_IFBLK - case S_IFBLK: - /* - * If -s has been specified, treat block special files - * like ordinary files. Otherwise, just report that they - * are block special files and go on to the next file. - */ - if ((ms->flags & MAGIC_DEVICES) != 0) { - ret = 0; - break; - } - if (mime) { - if (handle_mime(ms, mime, "blockdevice") == -1) - return -1; - } else if (silent) { - } else { -#ifdef HAVE_STRUCT_STAT_ST_RDEV -# ifdef dv_unit - if (file_printf(ms, "%sblock special (%d/%d/%d)", - COMMA, major(sb->st_rdev), dv_unit(sb->st_rdev), - dv_subunit(sb->st_rdev)) == -1) - return -1; -# else - if (file_printf(ms, "%sblock special (%ld/%ld)", - COMMA, (long)major(sb->st_rdev), - (long)minor(sb->st_rdev)) == -1) - return -1; + } + return 1; # endif -#else - if (file_printf(ms, "%sblock special", COMMA) == -1) - return -1; #endif - } - break; -#endif - /* TODO add code to handle V7 MUX and Blit MUX files */ + #ifdef S_IFIFO case S_IFIFO: if((ms->flags & MAGIC_DEVICES) != 0) @@ -292,92 +191,14 @@ #endif #ifdef S_IFLNK case S_IFLNK: - if ((nch = readlink(fn, buf, BUFSIZ-1)) <= 0) { + /* stat is used, if it made here then the link is broken */ if (ms->flags & MAGIC_ERROR) { - file_error(ms, errno, "unreadable symlink `%s'", - fn); + file_error(ms, errno, "unreadable symlink `%s'", fn); return -1; } - if (mime) { - if (handle_mime(ms, mime, "symlink") == -1) - return -1; - } else if (silent) { - } else if (file_printf(ms, - "%sunreadable symlink `%s' (%s)", COMMA, fn, - strerror(errno)) == -1) - return -1; - break; - } - buf[nch] = '\0'; /* readlink(2) does not do this */ - - /* If broken symlink, say so and quit early. */ -#ifdef __linux__ - /* - * linux procfs/devfs makes symlinks like pipe:[3515864880] - * that we can't stat their readlink output, so stat the - * original filename instead. - */ - if (stat(fn, &tstatbuf) < 0) - return bad_link(ms, errno, buf); -#else - if (*buf == '/') { - if (stat(buf, &tstatbuf) < 0) - return bad_link(ms, errno, buf); - } else { - char *tmp; - char buf2[BUFSIZ+BUFSIZ+4]; - - if ((tmp = CCAST(char *, strrchr(fn, '/'))) == NULL) { - tmp = buf; /* in current directory anyway */ - } else { - if (tmp - fn + 1 > BUFSIZ) { - if (ms->flags & MAGIC_ERROR) { - file_error(ms, 0, - "path too long: `%s'", buf); - return -1; - } - if (mime) { - if (handle_mime(ms, mime, - "x-path-too-long") == -1) - return -1; - } else if (silent) { - } else if (file_printf(ms, - "%spath too long: `%s'", COMMA, - fn) == -1) - return -1; - break; - } - /* take dir part */ - (void)strlcpy(buf2, fn, sizeof buf2); - buf2[tmp - fn + 1] = '\0'; - /* plus (rel) link */ - (void)strlcat(buf2, buf, sizeof buf2); - tmp = buf2; - } - if (stat(tmp, &tstatbuf) < 0) - return bad_link(ms, errno, buf); - } + return 1; #endif - /* Otherwise, handle it. */ - if ((ms->flags & MAGIC_SYMLINK) != 0) { - const char *p; - ms->flags &= MAGIC_SYMLINK; - p = magic_file(ms, buf); - ms->flags |= MAGIC_SYMLINK; - if (p == NULL) - return -1; - } else { /* just print what it points to */ - if (mime) { - if (handle_mime(ms, mime, "symlink") == -1) - return -1; - } else if (silent) { - } else if (file_printf(ms, "%ssymbolic link to %s", - COMMA, buf) == -1) - return -1; - } - break; -#endif #ifdef S_IFSOCK #ifndef __COHERENT__ case S_IFSOCK: diff -u libmagic.orig/funcs.c libmagic/funcs.c --- libmagic.orig/funcs.c 2023-07-27 21:40:12.000000000 +0200 +++ libmagic/funcs.c 2024-06-09 17:55:33.549243946 +0200 @@ -66,7 +66,7 @@ file_private void file_clearbuf(struct magic_set *ms) { - free(ms->o.buf); + efree(ms->o.buf); ms->o.buf = NULL; ms->o.blen = 0; } @@ -132,7 +132,7 @@ file_protected int file_vprintf(struct magic_set *ms, const char *fmt, va_list ap) { - int len; + size_t len; char *buf, *newstr; char tbuf[1024]; @@ -145,31 +145,25 @@ return -1; } - len = vasprintf(&buf, fmt, ap); - if (len < 0 || (size_t)len > 1024 || len + ms->o.blen > 1024 * 1024) { + len = vspprintf(&buf, 0, fmt, ap); + if (len > 1024 || len + ms->o.blen > 1024 * 1024) { size_t blen = ms->o.blen; - free(buf); + if (buf) efree(buf); file_clearbuf(ms); - file_error(ms, 0, "Output buffer space exceeded %d+%" + file_error(ms, 0, "Output buffer space exceeded %" SIZE_T_FORMAT "u+%" SIZE_T_FORMAT "u", len, blen); return -1; } if (ms->o.buf != NULL) { - len = asprintf(&newstr, "%s%s", ms->o.buf, buf); - free(buf); - if (len < 0) - goto out; - free(ms->o.buf); + len = spprintf(&newstr, 0, "%s%s", ms->o.buf, buf); + efree(buf); + efree(ms->o.buf); buf = newstr; } ms->o.buf = buf; ms->o.blen = len; return 0; -out: - file_clearbuf(ms); - file_error(ms, errno, "vasprintf failed"); - return -1; } file_protected int @@ -320,7 +314,7 @@ */ /*ARGSUSED*/ file_protected int -file_buffer(struct magic_set *ms, int fd, struct stat *st, +file_buffer(struct magic_set *ms, php_stream *stream, zend_stat_t *st, const char *inname __attribute__ ((__unused__)), const void *buf, size_t nb) { @@ -331,6 +325,19 @@ const char *ftype = NULL; char *rbuf = NULL; struct buffer b; + int fd = -1; + + if (stream) { +#ifdef _WIN64 + php_socket_t _fd = fd; +#else + int _fd; +#endif + int _ret = php_stream_cast(stream, PHP_STREAM_AS_FD, (void **)&_fd, 0); + if (SUCCESS == _ret) { + fd = (int)_fd; + } + } buffer_init(&b, fd, st, buf, nb); ms->mode = b.st.st_mode; @@ -363,7 +370,8 @@ } } #endif -#if HAVE_FORK + +#ifdef PHP_FILEINFO_UNCOMPRESS /* try compression stuff */ if ((ms->flags & MAGIC_NO_CHECK_COMPRESS) == 0) { m = file_zmagic(ms, &b, inname); @@ -447,7 +455,7 @@ rv = file_tryelf(ms, &b); rbuf = file_pop_buffer(ms, pb); if (rv == -1) { - free(rbuf); + efree(rbuf); rbuf = NULL; } if ((ms->flags & MAGIC_DEBUG) != 0) @@ -499,10 +507,10 @@ if (file_printf(ms, "%s", code_mime) == -1) rv = -1; } -#if HAVE_FORK +#ifdef PHP_FILEINFO_UNCOMPRESS done_encoding: #endif - free(rbuf); + efree(rbuf); buffer_fini(&b); if (rv) return rv; @@ -520,7 +528,7 @@ } file_clearbuf(ms); if (ms->o.pbuf) { - free(ms->o.pbuf); + efree(ms->o.pbuf); ms->o.pbuf = NULL; } ms->event_flags &= ~EVENT_HAD_ERR; @@ -558,7 +566,7 @@ return NULL; } psize = len * 4 + 1; - if ((pbuf = CAST(char *, realloc(ms->o.pbuf, psize))) == NULL) { + if ((pbuf = CAST(char *, erealloc(ms->o.pbuf, psize))) == NULL) { file_oomem(ms, psize); return NULL; } @@ -622,8 +630,8 @@ if (level >= ms->c.len) { len = (ms->c.len = 20 + level) * sizeof(*ms->c.li); ms->c.li = CAST(struct level_info *, (ms->c.li == NULL) ? - malloc(len) : - realloc(ms->c.li, len)); + emalloc(len) : + erealloc(ms->c.li, len)); if (ms->c.li == NULL) { file_oomem(ms, len); return -1; @@ -646,122 +654,38 @@ file_protected int file_replace(struct magic_set *ms, const char *pat, const char *rep) { - file_regex_t rx; - int rc, rv = -1; - - rc = file_regcomp(ms, &rx, pat, REG_EXTENDED); - if (rc == 0) { - regmatch_t rm; - int nm = 0; - while (file_regexec(ms, &rx, ms->o.buf, 1, &rm, 0) == 0) { - ms->o.buf[rm.rm_so] = '\0'; - if (file_printf(ms, "%s%s", rep, - rm.rm_eo != 0 ? ms->o.buf + rm.rm_eo : "") == -1) - goto out; - nm++; - } - rv = nm; + zend_string *pattern; + uint32_t opts = 0; + pcre_cache_entry *pce; + zend_string *res; + zend_string *repl; + size_t rep_cnt = 0; + + opts |= PCRE2_MULTILINE; + pattern = convert_libmagic_pattern((char*)pat, strlen(pat), opts); + if ((pce = pcre_get_compiled_regex_cache_ex(pattern, 0)) == NULL) { + zend_string_release(pattern); + rep_cnt = -1; + goto out; + } + zend_string_release(pattern); + + repl = zend_string_init(rep, strlen(rep), 0); + res = php_pcre_replace_impl(pce, NULL, ms->o.buf, strlen(ms->o.buf), repl, -1, &rep_cnt); + + zend_string_release_ex(repl, 0); + if (NULL == res) { + rep_cnt = -1; + goto out; } -out: - file_regfree(&rx); - return rv; -} -file_private int -check_regex(struct magic_set *ms, const char *pat) -{ - char sbuf[512]; - unsigned char oc = '\0'; - const char *p; + memcpy(ms->o.buf, ZSTR_VAL(res), ZSTR_LEN(res)); + ms->o.buf[ZSTR_LEN(res)] = '\0'; - for (p = pat; *p; p++) { - unsigned char c = *p; - // Avoid repetition - if (c == oc && strchr("?*+{", c) != NULL) { - size_t len = strlen(pat); - file_magwarn(ms, - "repetition-operator operand `%c' " - "invalid in regex `%s'", c, - file_printable(ms, sbuf, sizeof(sbuf), pat, len)); - return -1; - } - oc = c; - if (isprint(c) || isspace(c) || c == '\b' - || c == 0x8a) // XXX: apple magic fixme - continue; - size_t len = strlen(pat); - file_magwarn(ms, - "non-ascii characters in regex \\%#o `%s'", - c, file_printable(ms, sbuf, sizeof(sbuf), pat, len)); - return -1; - } - return 0; -} + zend_string_release_ex(res, 0); -file_protected int -file_regcomp(struct magic_set *ms file_locale_used, file_regex_t *rx, - const char *pat, int flags) -{ - if (check_regex(ms, pat) == -1) - return -1; - -#ifdef USE_C_LOCALE - locale_t old = uselocale(ms->c_lc_ctype); - assert(old != NULL); -#else - char old[1024]; - strlcpy(old, setlocale(LC_CTYPE, NULL), sizeof(old)); - (void)setlocale(LC_CTYPE, "C"); -#endif - int rc; - rc = regcomp(rx, pat, flags); - -#ifdef USE_C_LOCALE - uselocale(old); -#else - (void)setlocale(LC_CTYPE, old); -#endif - if (rc > 0 && (ms->flags & MAGIC_CHECK)) { - char errmsg[512], buf[512]; - - (void)regerror(rc, rx, errmsg, sizeof(errmsg)); - file_magerror(ms, "regex error %d for `%s', (%s)", rc, - file_printable(ms, buf, sizeof(buf), pat, strlen(pat)), - errmsg); - } - return rc; -} - -/*ARGSUSED*/ -file_protected int -file_regexec(struct magic_set *ms file_locale_used, file_regex_t *rx, - const char *str, size_t nmatch, regmatch_t* pmatch, int eflags) -{ -#ifdef USE_C_LOCALE - locale_t old = uselocale(ms->c_lc_ctype); - assert(old != NULL); -#else - char old[1024]; - strlcpy(old, setlocale(LC_CTYPE, NULL), sizeof(old)); - (void)setlocale(LC_CTYPE, "C"); -#endif - int rc; - /* XXX: force initialization because glibc does not always do this */ - if (nmatch != 0) - memset(pmatch, 0, nmatch * sizeof(*pmatch)); - rc = regexec(rx, str, nmatch, pmatch, eflags); -#ifdef USE_C_LOCALE - uselocale(old); -#else - (void)setlocale(LC_CTYPE, old); -#endif - return rc; -} - -file_protected void -file_regfree(file_regex_t *rx) -{ - regfree(rx); +out: + return rep_cnt; } file_protected file_pushbuf_t * @@ -772,7 +696,7 @@ if (ms->event_flags & EVENT_HAD_ERR) return NULL; - if ((pb = (CAST(file_pushbuf_t *, malloc(sizeof(*pb))))) == NULL) + if ((pb = (CAST(file_pushbuf_t *, emalloc(sizeof(*pb))))) == NULL) return NULL; pb->buf = ms->o.buf; @@ -792,8 +716,8 @@ char *rbuf; if (ms->event_flags & EVENT_HAD_ERR) { - free(pb->buf); - free(pb); + efree(pb->buf); + efree(pb); return NULL; } @@ -803,7 +727,7 @@ ms->o.blen = pb->blen; ms->offset = pb->offset; - free(pb); + efree(pb); return rbuf; } @@ -887,6 +811,7 @@ #endif } +#if 0 file_protected int file_pipe_closexec(int *fds) { @@ -904,6 +829,7 @@ return 0; #endif } +#endif file_protected int file_clear_closexec(int fd) { diff -u libmagic.orig/magic.c libmagic/magic.c --- libmagic.orig/magic.c 2023-07-27 21:33:24.000000000 +0200 +++ libmagic/magic.c 2024-06-09 00:31:40.347830761 +0200 @@ -25,11 +25,6 @@ * SUCH DAMAGE. */ -#ifdef WIN32 -#include -#include -#endif - #include "file.h" #ifndef lint @@ -39,7 +34,9 @@ #include "magic.h" #include +#ifdef HAVE_UNISTD_H #include +#endif #include #ifdef QUICK #include @@ -69,200 +66,18 @@ #endif #endif -file_private void close_and_restore(const struct magic_set *, const char *, int, - const struct stat *); -file_private int unreadable_info(struct magic_set *, mode_t, const char *); -file_private const char* get_default_magic(void); -#ifndef COMPILE_ONLY -file_private const char *file_or_fd(struct magic_set *, const char *, int); +#ifdef PHP_WIN32 +# undef S_IFLNK +# undef S_IFIFO #endif +file_private int unreadable_info(struct magic_set *, mode_t, const char *); +file_private const char *file_or_stream(struct magic_set *, const char *, php_stream *); + #ifndef STDIN_FILENO #define STDIN_FILENO 0 #endif -#ifdef WIN32 -/* HINSTANCE of this shared library. Needed for get_default_magic() */ -static HINSTANCE _w32_dll_instance = NULL; - -static void -_w32_append_path(char **hmagicpath, const char *fmt, ...) -{ - char *tmppath; - char *newpath; - va_list ap; - - va_start(ap, fmt); - if (vasprintf(&tmppath, fmt, ap) < 0) { - va_end(ap); - return; - } - va_end(ap); - - if (access(tmppath, R_OK) == -1) - goto out; - - if (*hmagicpath == NULL) { - *hmagicpath = tmppath; - return; - } - - if (asprintf(&newpath, "%s%c%s", *hmagicpath, PATHSEP, tmppath) < 0) - goto out; - - free(*hmagicpath); - free(tmppath); - *hmagicpath = newpath; - return; -out: - free(tmppath); -} - -static void -_w32_get_magic_relative_to(char **hmagicpath, HINSTANCE module) -{ - static const char *trypaths[] = { - "%s/share/misc/magic.mgc", - "%s/magic.mgc", - }; - LPSTR dllpath; - size_t sp; - - dllpath = calloc(MAX_PATH + 1, sizeof(*dllpath)); - - if (!GetModuleFileNameA(module, dllpath, MAX_PATH)) - goto out; - - PathRemoveFileSpecA(dllpath); - - if (module) { - char exepath[MAX_PATH]; - GetModuleFileNameA(NULL, exepath, MAX_PATH); - PathRemoveFileSpecA(exepath); - if (stricmp(exepath, dllpath) == 0) - goto out; - } - - sp = strlen(dllpath); - if (sp > 3 && stricmp(&dllpath[sp - 3], "bin") == 0) { - _w32_append_path(hmagicpath, - "%s/../share/misc/magic.mgc", dllpath); - goto out; - } - - for (sp = 0; sp < __arraycount(trypaths); sp++) - _w32_append_path(hmagicpath, trypaths[sp], dllpath); -out: - free(dllpath); -} - -#ifndef BUILD_AS_WINDOWS_STATIC_LIBARAY -/* Placate GCC by offering a sacrificial previous prototype */ -BOOL WINAPI DllMain(HINSTANCE, DWORD, LPVOID); - -BOOL WINAPI -DllMain(HINSTANCE hinstDLL, DWORD fdwReason, - LPVOID lpvReserved __attribute__((__unused__))) -{ - if (fdwReason == DLL_PROCESS_ATTACH) - _w32_dll_instance = hinstDLL; - return 1; -} -#endif -#endif - -file_private const char * -get_default_magic(void) -{ - static const char hmagic[] = "/.magic/magic.mgc"; - static char *default_magic; - char *home, *hmagicpath; - -#ifndef WIN32 - struct stat st; - - if (default_magic) { - free(default_magic); - default_magic = NULL; - } - if ((home = getenv("HOME")) == NULL) - return MAGIC; - - if (asprintf(&hmagicpath, "%s/.magic.mgc", home) < 0) - return MAGIC; - if (stat(hmagicpath, &st) == -1) { - free(hmagicpath); - if (asprintf(&hmagicpath, "%s/.magic", home) < 0) - return MAGIC; - if (stat(hmagicpath, &st) == -1) - goto out; - if (S_ISDIR(st.st_mode)) { - free(hmagicpath); - if (asprintf(&hmagicpath, "%s/%s", home, hmagic) < 0) - return MAGIC; - if (access(hmagicpath, R_OK) == -1) - goto out; - } - } - - if (asprintf(&default_magic, "%s:%s", hmagicpath, MAGIC) < 0) - goto out; - free(hmagicpath); - return default_magic; -out: - default_magic = NULL; - free(hmagicpath); - return MAGIC; -#else - hmagicpath = NULL; - - if (default_magic) { - free(default_magic); - default_magic = NULL; - } - - /* Before anything else, try to get a magic file from user HOME */ - if ((home = getenv("HOME")) != NULL) - _w32_append_path(&hmagicpath, "%s%s", home, hmagic); - - /* First, try to get a magic file from user-application data */ - if ((home = getenv("LOCALAPPDATA")) != NULL) - _w32_append_path(&hmagicpath, "%s%s", home, hmagic); - - /* Second, try to get a magic file from the user profile data */ - if ((home = getenv("USERPROFILE")) != NULL) - _w32_append_path(&hmagicpath, - "%s/Local Settings/Application Data%s", home, hmagic); - - /* Third, try to get a magic file from Common Files */ - if ((home = getenv("COMMONPROGRAMFILES")) != NULL) - _w32_append_path(&hmagicpath, "%s%s", home, hmagic); - - /* Fourth, try to get magic file relative to exe location */ - _w32_get_magic_relative_to(&hmagicpath, NULL); - - /* Fifth, try to get magic file relative to dll location */ - _w32_get_magic_relative_to(&hmagicpath, _w32_dll_instance); - - /* Avoid MAGIC constant - it likely points to a file within MSys tree */ - default_magic = hmagicpath; - return default_magic; -#endif -} - -file_public const char * -magic_getpath(const char *magicfile, int action) -{ - if (magicfile != NULL) - return magicfile; - - magicfile = getenv("MAGIC"); - if (magicfile != NULL) - return magicfile; - - return action == FILE_LOAD ? get_default_magic() : MAGIC; -} - file_public struct magic_set * magic_open(int flags) { @@ -321,21 +136,6 @@ return file_apprentice(ms, magicfile, FILE_LOAD); } -#ifndef COMPILE_ONLY -/* - * Install a set of compiled magic buffers. - */ -file_public int -magic_load_buffers(struct magic_set *ms, void **bufs, size_t *sizes, - size_t nbufs) -{ - if (ms == NULL) - return -1; - return buffer_apprentice(ms, RCAST(struct magic **, bufs), - sizes, nbufs); -} -#endif - file_public int magic_compile(struct magic_set *ms, const char *magicfile) { @@ -360,39 +160,6 @@ return file_apprentice(ms, magicfile, FILE_LIST); } -file_private void -close_and_restore(const struct magic_set *ms, const char *name, int fd, - const struct stat *sb) -{ - if (fd == STDIN_FILENO || name == NULL) - return; - (void) close(fd); - - if ((ms->flags & MAGIC_PRESERVE_ATIME) != 0) { - /* - * Try to restore access, modification times if read it. - * This is really *bad* because it will modify the status - * time of the file... And of course this will affect - * backup programs - */ -#ifdef HAVE_UTIMES - struct timeval utsbuf[2]; - (void)memset(utsbuf, 0, sizeof(utsbuf)); - utsbuf[0].tv_sec = sb->st_atime; - utsbuf[1].tv_sec = sb->st_mtime; - - (void) utimes(name, utsbuf); /* don't care if loses */ -#elif defined(HAVE_UTIME_H) || defined(HAVE_SYS_UTIME_H) - struct utimbuf utbuf; - - (void)memset(&utbuf, 0, sizeof(utbuf)); - utbuf.actime = sb->st_atime; - utbuf.modtime = sb->st_mtime; - (void) utime(name, &utbuf); /* don't care if loses */ -#endif - } -} - #ifndef COMPILE_ONLY /* @@ -403,7 +170,7 @@ { if (ms == NULL) return NULL; - return file_or_fd(ms, NULL, fd); + return file_or_stream(ms, NULL, NULL); } /* @@ -414,19 +181,25 @@ { if (ms == NULL) return NULL; - return file_or_fd(ms, inname, STDIN_FILENO); + return file_or_stream(ms, inname, NULL); +} + +file_public const char * +magic_stream(struct magic_set *ms, php_stream *stream) +{ + if (ms == NULL) + return NULL; + return file_or_stream(ms, NULL, stream); } file_private const char * -file_or_fd(struct magic_set *ms, const char *inname, int fd) +file_or_stream(struct magic_set *ms, const char *inname, php_stream *stream) { int rv = -1; unsigned char *buf; - struct stat sb; + zend_stat_t sb = {0}; ssize_t nbytes = 0; /* number of bytes read from a datafile */ - int ispipe = 0; - int okstat = 0; - off_t pos = CAST(off_t, -1); + int no_in_stream = 0; if (file_reset(ms, 1) == -1) goto out; @@ -436,7 +209,7 @@ * some overlapping space for matches near EOF */ #define SLOP (1 + sizeof(union VALUETYPE)) - if ((buf = CAST(unsigned char *, malloc(ms->bytes_max + SLOP))) == NULL) + if ((buf = CAST(unsigned char *, emalloc(ms->bytes_max + SLOP))) == NULL) return NULL; switch (file_fsmagic(ms, inname, &sb)) { @@ -449,96 +222,46 @@ goto done; } -#ifdef WIN32 - /* Place stdin in binary mode, so EOF (Ctrl+Z) doesn't stop early. */ - if (fd == STDIN_FILENO) - _setmode(STDIN_FILENO, O_BINARY); -#endif - if (inname != NULL) { - int flags = O_RDONLY|O_BINARY|O_NONBLOCK|O_CLOEXEC; - errno = 0; - if ((fd = open(inname, flags)) < 0) { - okstat = stat(inname, &sb) == 0; -#ifdef WIN32 - /* - * Can't stat, can't open. It may have been opened in - * fsmagic, so if the user doesn't have read permission, - * allow it to say so; otherwise an error was probably - * displayed in fsmagic. - */ - if (!okstat && errno == EACCES) { - sb.st_mode = S_IFBLK; - okstat = 1; - } -#endif - if (okstat && - unreadable_info(ms, sb.st_mode, inname) == -1) + errno = 0; + + if (inname && !stream) { + no_in_stream = 1; + stream = php_stream_open_wrapper((char *)inname, "rb", REPORT_ERRORS, NULL); + if (!stream) { + if (unreadable_info(ms, sb.st_mode, inname) == -1) goto done; - rv = 0; + rv = -1; goto done; } -#if O_CLOEXEC == 0 && defined(F_SETFD) - (void)fcntl(fd, F_SETFD, FD_CLOEXEC); -#endif } - if (fd != -1) { - okstat = fstat(fd, &sb) == 0; - if (okstat && S_ISFIFO(sb.st_mode)) - ispipe = 1; - if (inname == NULL) - pos = lseek(fd, CAST(off_t, 0), SEEK_CUR); + php_stream_statbuf ssb; + if (php_stream_stat(stream, &ssb) < 0) { + if (ms->flags & MAGIC_ERROR) { + file_error(ms, errno, "cannot stat `%s'", inname); + rv = -1; + goto done; + } } + memcpy(&sb, &ssb.sb, sizeof(zend_stat_t)); /* * try looking at the first ms->bytes_max bytes */ - if (ispipe) { - if (fd != -1) { - ssize_t r = 0; - - while ((r = sread(fd, RCAST(void *, &buf[nbytes]), - CAST(size_t, ms->bytes_max - nbytes), 1)) > 0) { - nbytes += r; - if (r < PIPE_BUF) break; - } - } - - if (nbytes == 0 && inname) { - /* We can not read it, but we were able to stat it. */ - if (unreadable_info(ms, sb.st_mode, inname) == -1) - goto done; - rv = 0; - goto done; - } - - } else if (fd != -1) { - /* Windows refuses to read from a big console buffer. */ - size_t howmany = -#ifdef WIN32 - _isatty(fd) ? 8 * 1024 : -#endif - ms->bytes_max; - if ((nbytes = read(fd, RCAST(void *, buf), howmany)) == -1) { - if (inname == NULL && fd != STDIN_FILENO) - file_error(ms, errno, "cannot read fd %d", fd); - else - file_error(ms, errno, "cannot read `%s'", - inname == NULL ? "/dev/stdin" : inname); - goto done; - } + if ((nbytes = php_stream_read(stream, (char *)buf, ms->bytes_max - nbytes)) < 0) { + file_error(ms, errno, "cannot read `%s'", inname); + goto done; } (void)memset(buf + nbytes, 0, SLOP); /* NUL terminate */ - if (file_buffer(ms, fd, okstat ? &sb : NULL, inname, buf, CAST(size_t, nbytes)) == -1) + if (file_buffer(ms, stream, &sb, inname, buf, CAST(size_t, nbytes)) == -1) goto done; rv = 0; done: - free(buf); - if (fd != -1) { - if (pos != CAST(off_t, -1)) - (void)lseek(fd, pos, SEEK_SET); - close_and_restore(ms, inname, fd, &sb); + efree(buf); + + if (no_in_stream && stream) { + php_stream_close(stream); } out: return rv == 0 ? file_getbuffer(ms) : NULL; @@ -556,7 +279,7 @@ * The main work is done here! * We have the file name and/or the data buffer to be identified. */ - if (file_buffer(ms, -1, NULL, NULL, buf, nb) == -1) { + if (file_buffer(ms, NULL, NULL, NULL, buf, nb) == -1) { return NULL; } return file_getbuffer(ms); diff -u libmagic.orig/magic.h libmagic/magic.h --- libmagic.orig/magic.h 2024-06-09 17:55:50.382419678 +0200 +++ libmagic/magic.h 2024-06-09 00:31:40.347830761 +0200 @@ -47,8 +47,6 @@ * extensions */ #define MAGIC_COMPRESS_TRANSP 0x2000000 /* Check inside compressed files * but not report compression */ -#define MAGIC_NO_COMPRESS_FORK 0x4000000 /* Don't allow decompression that - * needs to fork */ #define MAGIC_NODESC (MAGIC_EXTENSION|MAGIC_MIME|MAGIC_APPLE) #define MAGIC_NO_CHECK_COMPRESS 0x0001000 /* Don't check for compressed files */ @@ -77,7 +75,7 @@ MAGIC_NO_CHECK_TOKENS | \ MAGIC_NO_CHECK_ENCODING | \ MAGIC_NO_CHECK_JSON | \ - MAGIC_NO_CHECK_SIMH | \ + MAGIC_NO_CHECK_SIMH | \ 0 \ ) @@ -101,11 +99,11 @@ b\20no_check_elf\0\ b\21no_check_text\0\ b\22no_check_cdf\0\ -b\23no_check_csv\0\ +b\23no_check_reserved0\0\ b\24no_check_tokens\0\ b\25no_check_encoding\0\ b\26no_check_json\0\ -b\27no_check_simh\0\ +b\27no_check_reserved2\0\ b\30extension\0\ b\31transp_compression\0\ " @@ -130,6 +128,7 @@ const char *magic_getpath(const char *, int); const char *magic_file(magic_t, const char *); +const char *magic_stream(magic_t, php_stream *); const char *magic_descriptor(magic_t, int); const char *magic_buffer(magic_t, const void *, size_t); @@ -154,7 +153,8 @@ #define MAGIC_PARAM_REGEX_MAX 5 #define MAGIC_PARAM_BYTES_MAX 6 #define MAGIC_PARAM_ENCODING_MAX 7 -#define MAGIC_PARAM_ELF_SHSIZE_MAX 8 +#define MAGIC_PARAM_ELF_SHSIZE_MAX 8 +#define MAGIC_PARAM_MAGWARN_MAX 9 int magic_setparam(magic_t, int, const void *); int magic_getparam(magic_t, int, void *); diff -u libmagic.orig/print.c libmagic/print.c --- libmagic.orig/print.c 2023-07-27 20:04:45.000000000 +0200 +++ libmagic/print.c 2024-06-09 00:31:40.347830761 +0200 @@ -73,7 +73,7 @@ if (m->mask_op & FILE_OPINVERSE) (void) fputc('~', stderr); - if (IS_STRING(m->type)) { + if (IS_LIBMAGIC_STRING(m->type)) { if (m->str_flags) { (void) fputc('/', stderr); if (m->str_flags & STRING_COMPACT_WHITESPACE) @@ -246,18 +246,18 @@ file_magwarn(struct magic_set *ms, const char *f, ...) { va_list va; + char *expanded_format = NULL; + int expanded_len; - /* cuz we use stdout for most, stderr here */ - (void) fflush(stdout); - - if (ms && ms->file) - (void) fprintf(stderr, "%s, %lu: ", ms->file, - CAST(unsigned long, ms->line)); - (void) fprintf(stderr, "Warning: "); va_start(va, f); - (void) vfprintf(stderr, f, va); + expanded_len = vasprintf(&expanded_format, f, va); va_end(va); - (void) fputc('\n', stderr); + + if (expanded_len >= 0 && expanded_format) { + php_error_docref(NULL, E_WARNING, "%s", expanded_format); + + free(expanded_format); + } } file_protected const char * @@ -289,13 +289,13 @@ goto out; if (flags & FILE_T_LOCAL) { - tm = localtime_r(&t, &tmz); + tm = php_localtime_r(&t, &tmz); } else { - tm = gmtime_r(&t, &tmz); + tm = php_gmtime_r(&t, &tmz); } if (tm == NULL) goto out; - pp = asctime_r(tm, buf); + pp = php_asctime_r(tm, buf); if (pp == NULL) goto out; diff -u libmagic.orig/readcdf.c libmagic/readcdf.c --- libmagic.orig/readcdf.c 2023-02-09 18:43:53.000000000 +0100 +++ libmagic/readcdf.c 2024-06-09 00:31:40.347830761 +0200 @@ -31,7 +31,9 @@ #include #include +#ifdef HAVE_UNISTD_H #include +#endif #include #include #include @@ -100,10 +102,6 @@ if (clsid[0] == cv[i].clsid[0] && clsid[1] == cv[i].clsid[1]) return cv[i].mime; } -#ifdef CDF_DEBUG - fprintf(stderr, "unknown mime %" PRIx64 ", %" PRIx64 "\n", clsid[0], - clsid[1]); -#endif return NULL; } @@ -112,35 +110,24 @@ { size_t i; const char *rv = NULL; -#ifdef USE_C_LOCALE - locale_t old_lc_ctype, c_lc_ctype; + char *vbuf_lower; - c_lc_ctype = newlocale(LC_CTYPE_MASK, "C", 0); - assert(c_lc_ctype != NULL); - old_lc_ctype = uselocale(c_lc_ctype); - assert(old_lc_ctype != NULL); -#else - char *old_lc_ctype = setlocale(LC_CTYPE, NULL); - assert(old_lc_ctype != NULL); - old_lc_ctype = strdup(old_lc_ctype); - assert(old_lc_ctype != NULL); - (void)setlocale(LC_CTYPE, "C"); -#endif - for (i = 0; nv[i].pattern != NULL; i++) - if (strcasestr(vbuf, nv[i].pattern) != NULL) { + vbuf_lower = zend_str_tolower_dup(vbuf, strlen(vbuf)); + for (i = 0; nv[i].pattern != NULL; i++) { + char *pattern_lower; + int found; + + pattern_lower = zend_str_tolower_dup(nv[i].pattern, strlen(nv[i].pattern)); + found = (strstr(vbuf_lower, pattern_lower) != NULL); + efree(pattern_lower); + + if (found) { rv = nv[i].mime; break; } -#ifdef CDF_DEBUG - fprintf(stderr, "unknown app %s\n", vbuf); -#endif -#ifdef USE_C_LOCALE - (void)uselocale(old_lc_ctype); - freelocale(c_lc_ctype); -#else - (void)setlocale(LC_CTYPE, old_lc_ctype); - free(old_lc_ctype); -#endif + } + + efree(vbuf_lower); return rv; } @@ -156,6 +143,8 @@ const char *s, *e; int len; + memset(&ts, 0, sizeof(ts)); + if (!NOTMIME(ms) && root_storage) str = cdf_clsid_to_mime(root_storage->d_storage_uuid, clsid2mime); @@ -282,10 +271,10 @@ if (file_printf(ms, "%s%s", cdf_u16tos8(buf, ce[i].ce_namlen, ce[i].ce_name), i == cat->cat_num - 1 ? "]" : ", ") == -1) { - free(cat); + efree(cat); return -1; } - free(cat); + efree(cat); } else if (ms->flags & MAGIC_MIME_TYPE) { if (file_printf(ms, "application/CDFV2") == -1) return -1; @@ -346,7 +335,7 @@ } m = cdf_file_property_info(ms, info, count, root_storage); - free(info); + efree(info); return m == -1 ? -2 : m; } @@ -656,11 +645,11 @@ cdf_zero_stream(&scn); cdf_zero_stream(&sst); out3: - free(dir.dir_tab); + efree(dir.dir_tab); out2: - free(ssat.sat_tab); + efree(ssat.sat_tab); out1: - free(sat.sat_tab); + efree(sat.sat_tab); out0: /* If we handled it already, return */ if (i != -1) diff -u libmagic.orig/softmagic.c libmagic/softmagic.c --- libmagic.orig/softmagic.c 2023-07-27 21:40:12.000000000 +0200 +++ libmagic/softmagic.c 2024-06-09 00:31:40.347830761 +0200 @@ -45,7 +45,7 @@ #include #include "der.h" -file_private int match(struct magic_set *, struct magic *, file_regex_t **, size_t, +file_private int match(struct magic_set *, struct magic *, size_t, const struct buffer *, size_t, int, int, int, uint16_t *, uint16_t *, int *, int *, int *, int *, int *); file_private int mget(struct magic_set *, struct magic *, const struct buffer *, @@ -54,7 +54,7 @@ uint16_t *, int *, int *, int *, int *, int *); file_private int msetoffset(struct magic_set *, struct magic *, struct buffer *, const struct buffer *, size_t, unsigned int); -file_private int magiccheck(struct magic_set *, struct magic *, file_regex_t **); +file_private int magiccheck(struct magic_set *, struct magic *); file_private int mprint(struct magic_set *, struct magic *); file_private int moffset(struct magic_set *, struct magic *, const struct buffer *, int32_t *); @@ -133,7 +133,7 @@ } for (ml = ms->mlist[0]->next; ml != ms->mlist[0]; ml = ml->next) { - int ret = match(ms, ml->magic, ml->magic_rxcomp, ml->nmagic, b, + int ret = match(ms, ml->magic, ml->nmagic, b, 0, mode, text, 0, indir_count, name_count, &printed_something, &need_separator, &firstline, NULL, NULL); @@ -153,8 +153,8 @@ return rv; } -#define FILE_FMTDEBUG -#ifdef FILE_FMTDEBUG + +#if defined(FILE_FMTDEBUG) && defined(HAVE_FMTCHECK) #define F(a, b, c) file_fmtcheck((a), (b), (c), __FILE__, __LINE__) file_private const char * __attribute__((__format_arg__(3))) @@ -173,10 +173,14 @@ " with `%s'", file, line, desc, def); return ptr; } -#else +#elif defined(HAVE_FMTCHECK) #define F(a, b, c) fmtcheck((b), (c)) +#else +#define F(a, b, c) ((b)) #endif +/* NOTE this function has been kept an the state of 5.39 for BC. Observe + * further as the upgrade to 5.41 or above goes. */ /* * Go through the whole list, stopping if you find a match. Process all * the continuations of that match before returning. @@ -205,7 +209,7 @@ * so that higher-level continuations are processed. */ file_private int -match(struct magic_set *ms, struct magic *magic, file_regex_t **magic_rxcomp, +match(struct magic_set *ms, struct magic *magic, size_t nmagic, const struct buffer *b, size_t offset, int mode, int text, int flip, uint16_t *indir_count, uint16_t *name_count, int *printed_something, int *need_separator, int *firstline, @@ -233,10 +237,9 @@ for (magindex = 0; magindex < nmagic; magindex++) { int flush = 0; struct magic *m = &magic[magindex]; - file_regex_t **m_rxcomp = &magic_rxcomp[magindex]; if (m->type != FILE_NAME) - if ((IS_STRING(m->type) && + if ((IS_LIBMAGIC_STRING(m->type) && #define FLT (STRING_BINTEST | STRING_TEXTTEST) ((text && (m->str_flags & FLT) == STRING_BINTEST) || (!text && (m->str_flags & FLT) == STRING_TEXTTEST))) || @@ -272,7 +275,7 @@ *returnval = 1; } - switch (magiccheck(ms, m, m_rxcomp)) { + switch (magiccheck(ms, m)) { case -1: return -1; case 0: @@ -334,7 +337,6 @@ while (magindex + 1 < nmagic && magic[magindex + 1].cont_level != 0) { m = &magic[++magindex]; - m_rxcomp = &magic_rxcomp[magindex]; ms->line = m->lineno; /* for messages */ if (cont_level < m->cont_level) @@ -388,7 +390,7 @@ break; } - switch (flush ? 1 : magiccheck(ms, m, m_rxcomp)) { + switch (flush ? 1 : magiccheck(ms, m)) { case -1: return -1; case 0: @@ -487,19 +489,25 @@ file_private int check_fmt(struct magic_set *ms, const char *fmt) { - file_regex_t rx; - int rc, rv = -1; - const char* pat = "%[-0-9\\.]*s"; + pcre_cache_entry *pce; + int rv = -1; + zend_string *pattern; if (strchr(fmt, '%') == NULL) return 0; - rc = file_regcomp(ms, &rx, pat, REG_EXTENDED|REG_NOSUB); - if (rc == 0) { - rc = file_regexec(ms, &rx, fmt, 0, 0, 0); - rv = !rc; + pattern = ZSTR_INIT_LITERAL("~%[-0-9\\.]*s~", 0); + if ((pce = pcre_get_compiled_regex_cache_ex(pattern, 0)) == NULL) { + rv = -1; + } else { + pcre2_code *re = php_pcre_pce_re(pce); + pcre2_match_data *match_data = php_pcre_create_match_data(0, re); + if (match_data) { + rv = pcre2_match(re, (PCRE2_SPTR)fmt, strlen(fmt), 0, 0, match_data, php_pcre_mctx()) > 0; + php_pcre_free_match_data(match_data); + } } - file_regfree(&rx); + zend_string_release_ex(pattern, 0); return rv; } @@ -517,7 +525,7 @@ for (len = 0; len < n && str[len]; len++) continue; - if ((copy = CAST(char *, malloc(len + 1))) == NULL) + if ((copy = CAST(char *, emalloc(len + 1))) == NULL) return NULL; (void)memcpy(copy, str, len); copy[len] = '\0'; @@ -767,7 +775,7 @@ char *cp, *scp; int rval; - cp = strndup(RCAST(const char *, ms->search.s), + cp = estrndup(RCAST(const char *, ms->search.s), ms->search.rm_len); if (cp == NULL) { file_oomem(ms, ms->search.rm_len); @@ -777,7 +785,7 @@ rval = file_printf(ms, F(ms, desc, "%s"), file_printable(ms, sbuf, sizeof(sbuf), scp, ms->search.rm_len)); - free(cp); + efree(cp); if (rval == -1) return -1; @@ -1565,7 +1573,7 @@ size_t len; *c = ms->c; len = c->len * sizeof(*c->li); - ms->c.li = CAST(struct level_info *, malloc(len)); + ms->c.li = CAST(struct level_info *, emalloc(len)); if (ms->c.li == NULL) { ms->c = *c; return -1; @@ -1577,7 +1585,7 @@ file_private void restore_cont(struct magic_set *ms, struct cont *c) { - free(ms->c.li); + efree(ms->c.li); ms->c = *c; } @@ -1894,7 +1902,7 @@ for (mlp = ms->mlist[0]->next; mlp != ms->mlist[0]; mlp = mlp->next) { - if ((rv = match(ms, mlp->magic, mlp->magic_rxcomp, + if ((rv = match(ms, mlp->magic, mlp->nmagic, &bb, 0, BINTEST, text, 0, indir_count, name_count, printed_something, need_separator, firstline, NULL, NULL)) != 0) @@ -1913,15 +1921,15 @@ if ((ms->flags & MAGIC_NODESC) == 0 && file_printf(ms, F(ms, m->desc, "%u"), offset) == -1) { - free(rbuf); + if (rbuf) efree(rbuf); return -1; } if (file_printf(ms, "%s", rbuf) == -1) { - free(rbuf); + if (rbuf) efree(rbuf); return -1; } } - free(rbuf); + if (rbuf) efree(rbuf); return rv; case FILE_USE: @@ -1948,7 +1956,7 @@ nfound_match = 0; (*name_count)++; eoffset = ms->eoffset; - rv = match(ms, ml.magic, ml.magic_rxcomp, ml.nmagic, b, + rv = match(ms, ml.magic, ml.nmagic, b, offset + o, mode, text, flip, indir_count, name_count, printed_something, need_separator, firstline, returnval, &nfound_match); @@ -2027,11 +2035,13 @@ } else if ((flags & STRING_COMPACT_WHITESPACE) && isspace(*a)) { + /* XXX Dirty. The data and the pattern is what is causing this. + Revert _i for the next port and see if it still matters. */ + uint32_t _i = 0; a++; - if (isspace(*b)) { - b++; + if (isspace(*b++)) { if (!isspace(*a)) - while (b < eb && isspace(*b)) + while (EXPECTED(_i++ < 2048) && b < eb && isspace(*b)) b++; } else { @@ -2071,29 +2081,8 @@ return file_strncmp(a, b, len, maxlen, flags); } -file_private file_regex_t * -alloc_regex(struct magic_set *ms, struct magic *m) -{ - int rc; - file_regex_t *rx = CAST(file_regex_t *, malloc(sizeof(*rx))); - - if (rx == NULL) { - file_error(ms, errno, "can't allocate %" SIZE_T_FORMAT - "u bytes", sizeof(*rx)); - return NULL; - } - - rc = file_regcomp(ms, rx, m->value.s, REG_EXTENDED | REG_NEWLINE | - ((m->str_flags & STRING_IGNORE_CASE) ? REG_ICASE : 0)); - if (rc == 0) - return rx; - - free(rx); - return NULL; -} - file_private int -magiccheck(struct magic_set *ms, struct magic *m, file_regex_t **m_cache) +magiccheck(struct magic_set *ms, struct magic *m) { uint64_t l = m->value.q; uint64_t v; @@ -2247,28 +2236,14 @@ slen = MIN(m->vallen, sizeof(m->value.s)); l = 0; v = 0; - if ((ms->flags & MAGIC_DEBUG) != 0) { - size_t xlen = ms->search.s_len > 100 ? 100 - : ms->search.s_len; - - fprintf(stderr, "search: ["); - file_showstr(stderr, ms->search.s, xlen); - fprintf(stderr, "%s] for [", ms->search.s_len == xlen - ? "" : "..."); - file_showstr(stderr, m->value.s, slen); - } #ifdef HAVE_MEMMEM if (slen > 0 && m->str_flags == 0) { const char *found; idx = m->str_range + slen; if (m->str_range == 0 || ms->search.s_len < idx) idx = ms->search.s_len; - found = CAST(const char *, memmem(ms->search.s, idx, - m->value.s, slen)); - if ((ms->flags & MAGIC_DEBUG) != 0) { - fprintf(stderr, "] %sfound\n", - found ? "" : "not "); - } + found = CAST(const char *, php_memnstr(ms->search.s, + m->value.s, slen, ms->search.s + idx)); if (!found) { v = 1; break; @@ -2294,61 +2269,79 @@ break; } } - if ((ms->flags & MAGIC_DEBUG) != 0) { - fprintf(stderr, "] %sfound\n", v == 0 ? "" : "not "); - } break; } case FILE_REGEX: { - int rc; - file_regex_t *rx = *m_cache; - const char *search; - regmatch_t pmatch; - size_t slen = ms->search.s_len; - char *copy; + zend_string *pattern; + uint32_t options = 0; + pcre_cache_entry *pce; - if (ms->search.s == NULL) - return 0; + options |= PCRE2_MULTILINE; - if (rx == NULL) { - rx = *m_cache = alloc_regex(ms, m); - if (rx == NULL) - return -1; + if (m->str_flags & STRING_IGNORE_CASE) { + options |= PCRE2_CASELESS; } - l = 0; - if (slen != 0) { - copy = CAST(char *, malloc(slen)); - if (copy == NULL) { - file_error(ms, errno, - "can't allocate %" SIZE_T_FORMAT "u bytes", - slen); - return -1; - } - memcpy(copy, ms->search.s, slen); - copy[--slen] = '\0'; - search = copy; - } else { - search = CCAST(char *, ""); - copy = NULL; - } - rc = file_regexec(ms, rx, RCAST(const char *, search), - 1, &pmatch, 0); - free(copy); - switch (rc) { - case 0: - ms->search.s += CAST(int, pmatch.rm_so); - ms->search.offset += CAST(size_t, pmatch.rm_so); - ms->search.rm_len = CAST(size_t, - pmatch.rm_eo - pmatch.rm_so); - v = 0; - break; - case REG_NOMATCH: - v = 1; - break; + pattern = convert_libmagic_pattern((char *)m->value.s, m->vallen, options); - default: + l = v = 0; + if ((pce = pcre_get_compiled_regex_cache(pattern)) == NULL) { + zend_string_release(pattern); return -1; + } else { + /* pce now contains the compiled regex */ + zval retval; + zval subpats; + zend_string *haystack; + + ZVAL_NULL(&retval); + ZVAL_NULL(&subpats); + + /* Cut the search len from haystack, equals to REG_STARTEND */ + haystack = zend_string_init(ms->search.s, ms->search.s_len, 0); + + /* match v = 0, no match v = 1 */ + php_pcre_match_impl(pce, haystack, &retval, &subpats, 0, PREG_OFFSET_CAPTURE, 0); + /* Free haystack */ + zend_string_release(haystack); + + if (Z_LVAL(retval) < 0) { + zval_ptr_dtor(&subpats); + zend_string_release(pattern); + return -1; + } else if ((Z_LVAL(retval) > 0) && (Z_TYPE(subpats) == IS_ARRAY)) { + /* Need to fetch global match which equals pmatch[0] */ + zval *pzval; + HashTable *ht = Z_ARRVAL(subpats); + if ((pzval = zend_hash_index_find(ht, 0)) != NULL && Z_TYPE_P(pzval) == IS_ARRAY) { + /* If everything goes according to the master plan + tmpcopy now contains two elements: + 0 = the match + 1 = starting position of the match */ + zval *match, *offset; + if ((match = zend_hash_index_find(Z_ARRVAL_P(pzval), 0)) && + (offset = zend_hash_index_find(Z_ARRVAL_P(pzval), 1))) { + if (Z_TYPE_P(match) != IS_STRING && Z_TYPE_P(offset) != IS_LONG) { + goto error_out; + } + ms->search.s += Z_LVAL_P(offset); /* this is where the match starts */ + ms->search.offset += Z_LVAL_P(offset); /* this is where the match starts as size_t */ + ms->search.rm_len = Z_STRLEN_P(match) /* This is the length of the matched pattern */; + v = 0; + } else { + goto error_out; + } + } else { +error_out: + zval_ptr_dtor(&subpats); + zend_string_release(pattern); + return -1; + } + } else { + v = 1; + } + zval_ptr_dtor(&subpats); + zend_string_release(pattern); } break; }