tests: ensure fsck fails on corrupt packfiles
[gitweb.git] / sha1_file.c
index 59a4ed2ed32336b41ab8f3b1d4aca30f045aa084..b60ae15f7068c157df6407933ea7ee94a53f0640 100644 (file)
@@ -7,6 +7,7 @@
  * creation etc.
  */
 #include "cache.h"
+#include "config.h"
 #include "string-list.h"
 #include "lockfile.h"
 #include "delta.h"
@@ -610,8 +611,7 @@ char *compute_alternate_path(const char *path, struct strbuf *err)
 
 out:
        if (seen_error) {
-               free(ref_git);
-               ref_git = NULL;
+               FREE_AND_NULL(ref_git);
        }
 
        return ref_git;
@@ -1684,14 +1684,14 @@ int git_open_cloexec(const char *name, int flags)
                fd = open(name, flags | o_cloexec);
        }
 
-#if defined(F_GETFL) && defined(F_SETFL) && defined(FD_CLOEXEC)
+#if defined(F_GETFD) && defined(F_SETFD) && defined(FD_CLOEXEC)
        {
                static int fd_cloexec = FD_CLOEXEC;
 
                if (!o_cloexec && 0 <= fd && fd_cloexec) {
                        /* Opened w/o O_CLOEXEC?  try with fcntl(2) to add it */
-                       int flags = fcntl(fd, F_GETFL);
-                       if (fcntl(fd, F_SETFL, flags | fd_cloexec))
+                       int flags = fcntl(fd, F_GETFD);
+                       if (fcntl(fd, F_SETFD, flags | fd_cloexec))
                                fd_cloexec = 0;
                }
        }
@@ -1964,7 +1964,7 @@ static int parse_sha1_header_extended(const char *hdr, struct object_info *oi,
         * we're obtaining the type using '--allow-unknown-type'
         * option.
         */
-       if ((flags & LOOKUP_UNKNOWN_OBJECT) && (type < 0))
+       if ((flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE) && (type < 0))
                type = 0;
        else if (type < 0)
                die("invalid object type");
@@ -2002,20 +2002,7 @@ int parse_sha1_header(const char *hdr, unsigned long *sizep)
        struct object_info oi = OBJECT_INFO_INIT;
 
        oi.sizep = sizep;
-       return parse_sha1_header_extended(hdr, &oi, LOOKUP_REPLACE_OBJECT);
-}
-
-static void *unpack_sha1_file(void *map, unsigned long mapsize, enum object_type *type, unsigned long *size, const unsigned char *sha1)
-{
-       int ret;
-       git_zstream stream;
-       char hdr[8192];
-
-       ret = unpack_sha1_header(&stream, map, mapsize, hdr, sizeof(hdr));
-       if (ret < Z_OK || (*type = parse_sha1_header(hdr, size)) < 0)
-               return NULL;
-
-       return unpack_sha1_rest(&stream, hdr, *size, sha1);
+       return parse_sha1_header_extended(hdr, &oi, 0);
 }
 
 unsigned long get_size_from_delta(struct packed_git *p,
@@ -2239,107 +2226,6 @@ static enum object_type packed_to_object_type(struct packed_git *p,
        goto out;
 }
 
-int packed_object_info(struct packed_git *p, off_t obj_offset,
-                      struct object_info *oi)
-{
-       struct pack_window *w_curs = NULL;
-       unsigned long size;
-       off_t curpos = obj_offset;
-       enum object_type type;
-
-       /*
-        * We always get the representation type, but only convert it to
-        * a "real" type later if the caller is interested.
-        */
-       type = unpack_object_header(p, &w_curs, &curpos, &size);
-
-       if (oi->sizep) {
-               if (type == OBJ_OFS_DELTA || type == OBJ_REF_DELTA) {
-                       off_t tmp_pos = curpos;
-                       off_t base_offset = get_delta_base(p, &w_curs, &tmp_pos,
-                                                          type, obj_offset);
-                       if (!base_offset) {
-                               type = OBJ_BAD;
-                               goto out;
-                       }
-                       *oi->sizep = get_size_from_delta(p, &w_curs, tmp_pos);
-                       if (*oi->sizep == 0) {
-                               type = OBJ_BAD;
-                               goto out;
-                       }
-               } else {
-                       *oi->sizep = size;
-               }
-       }
-
-       if (oi->disk_sizep) {
-               struct revindex_entry *revidx = find_pack_revindex(p, obj_offset);
-               *oi->disk_sizep = revidx[1].offset - obj_offset;
-       }
-
-       if (oi->typep) {
-               *oi->typep = packed_to_object_type(p, obj_offset, type, &w_curs, curpos);
-               if (*oi->typep < 0) {
-                       type = OBJ_BAD;
-                       goto out;
-               }
-       }
-
-       if (oi->delta_base_sha1) {
-               if (type == OBJ_OFS_DELTA || type == OBJ_REF_DELTA) {
-                       const unsigned char *base;
-
-                       base = get_delta_base_sha1(p, &w_curs, curpos,
-                                                  type, obj_offset);
-                       if (!base) {
-                               type = OBJ_BAD;
-                               goto out;
-                       }
-
-                       hashcpy(oi->delta_base_sha1, base);
-               } else
-                       hashclr(oi->delta_base_sha1);
-       }
-
-out:
-       unuse_pack(&w_curs);
-       return type;
-}
-
-static void *unpack_compressed_entry(struct packed_git *p,
-                                   struct pack_window **w_curs,
-                                   off_t curpos,
-                                   unsigned long size)
-{
-       int st;
-       git_zstream stream;
-       unsigned char *buffer, *in;
-
-       buffer = xmallocz_gently(size);
-       if (!buffer)
-               return NULL;
-       memset(&stream, 0, sizeof(stream));
-       stream.next_out = buffer;
-       stream.avail_out = size + 1;
-
-       git_inflate_init(&stream);
-       do {
-               in = use_pack(p, w_curs, curpos, &stream.avail_in);
-               stream.next_in = in;
-               st = git_inflate(&stream, Z_FINISH);
-               if (!stream.avail_out)
-                       break; /* the payload is larger than it should be */
-               curpos += stream.next_in - in;
-       } while (st == Z_OK || st == Z_BUF_ERROR);
-       git_inflate_end(&stream);
-       if ((st != Z_STREAM_END) || stream.total_out != size) {
-               free(buffer);
-               return NULL;
-       }
-
-       return buffer;
-}
-
 static struct hashmap delta_base_cache;
 static size_t delta_base_cached;
 
@@ -2389,7 +2275,8 @@ static int delta_base_cache_key_eq(const struct delta_base_cache_key *a,
        return a->p == b->p && a->base_offset == b->base_offset;
 }
 
-static int delta_base_cache_hash_cmp(const void *va, const void *vb,
+static int delta_base_cache_hash_cmp(const void *unused_cmp_data,
+                                    const void *va, const void *vb,
                                     const void *vkey)
 {
        const struct delta_base_cache_entry *a = va, *b = vb;
@@ -2427,8 +2314,10 @@ static void *cache_or_unpack_entry(struct packed_git *p, off_t base_offset,
        if (!ent)
                return unpack_entry(p, base_offset, type, base_size);
 
-       *type = ent->type;
-       *base_size = ent->size;
+       if (type)
+               *type = ent->type;
+       if (base_size)
+               *base_size = ent->size;
        return xmemdupz(ent->data, ent->size);
 }
 
@@ -2472,11 +2361,128 @@ static void add_delta_base_cache(struct packed_git *p, off_t base_offset,
        list_add_tail(&ent->lru, &delta_base_cache_lru);
 
        if (!delta_base_cache.cmpfn)
-               hashmap_init(&delta_base_cache, delta_base_cache_hash_cmp, 0);
+               hashmap_init(&delta_base_cache, delta_base_cache_hash_cmp, NULL, 0);
        hashmap_entry_init(ent, pack_entry_hash(p, base_offset));
        hashmap_add(&delta_base_cache, ent);
 }
 
+int packed_object_info(struct packed_git *p, off_t obj_offset,
+                      struct object_info *oi)
+{
+       struct pack_window *w_curs = NULL;
+       unsigned long size;
+       off_t curpos = obj_offset;
+       enum object_type type;
+
+       /*
+        * We always get the representation type, but only convert it to
+        * a "real" type later if the caller is interested.
+        */
+       if (oi->contentp) {
+               *oi->contentp = cache_or_unpack_entry(p, obj_offset, oi->sizep,
+                                                     &type);
+               if (!*oi->contentp)
+                       type = OBJ_BAD;
+       } else {
+               type = unpack_object_header(p, &w_curs, &curpos, &size);
+       }
+
+       if (!oi->contentp && oi->sizep) {
+               if (type == OBJ_OFS_DELTA || type == OBJ_REF_DELTA) {
+                       off_t tmp_pos = curpos;
+                       off_t base_offset = get_delta_base(p, &w_curs, &tmp_pos,
+                                                          type, obj_offset);
+                       if (!base_offset) {
+                               type = OBJ_BAD;
+                               goto out;
+                       }
+                       *oi->sizep = get_size_from_delta(p, &w_curs, tmp_pos);
+                       if (*oi->sizep == 0) {
+                               type = OBJ_BAD;
+                               goto out;
+                       }
+               } else {
+                       *oi->sizep = size;
+               }
+       }
+
+       if (oi->disk_sizep) {
+               struct revindex_entry *revidx = find_pack_revindex(p, obj_offset);
+               *oi->disk_sizep = revidx[1].offset - obj_offset;
+       }
+
+       if (oi->typep || oi->typename) {
+               enum object_type ptot;
+               ptot = packed_to_object_type(p, obj_offset, type, &w_curs,
+                                            curpos);
+               if (oi->typep)
+                       *oi->typep = ptot;
+               if (oi->typename) {
+                       const char *tn = typename(ptot);
+                       if (tn)
+                               strbuf_addstr(oi->typename, tn);
+               }
+               if (ptot < 0) {
+                       type = OBJ_BAD;
+                       goto out;
+               }
+       }
+
+       if (oi->delta_base_sha1) {
+               if (type == OBJ_OFS_DELTA || type == OBJ_REF_DELTA) {
+                       const unsigned char *base;
+
+                       base = get_delta_base_sha1(p, &w_curs, curpos,
+                                                  type, obj_offset);
+                       if (!base) {
+                               type = OBJ_BAD;
+                               goto out;
+                       }
+
+                       hashcpy(oi->delta_base_sha1, base);
+               } else
+                       hashclr(oi->delta_base_sha1);
+       }
+
+out:
+       unuse_pack(&w_curs);
+       return type;
+}
+
+static void *unpack_compressed_entry(struct packed_git *p,
+                                   struct pack_window **w_curs,
+                                   off_t curpos,
+                                   unsigned long size)
+{
+       int st;
+       git_zstream stream;
+       unsigned char *buffer, *in;
+
+       buffer = xmallocz_gently(size);
+       if (!buffer)
+               return NULL;
+       memset(&stream, 0, sizeof(stream));
+       stream.next_out = buffer;
+       stream.avail_out = size + 1;
+
+       git_inflate_init(&stream);
+       do {
+               in = use_pack(p, w_curs, curpos, &stream.avail_in);
+               stream.next_in = in;
+               st = git_inflate(&stream, Z_FINISH);
+               if (!stream.avail_out)
+                       break; /* the payload is larger than it should be */
+               curpos += stream.next_in - in;
+       } while (st == Z_OK || st == Z_BUF_ERROR);
+       git_inflate_end(&stream);
+       if ((st != Z_STREAM_END) || stream.total_out != size) {
+               free(buffer);
+               return NULL;
+       }
+
+       return buffer;
+}
+
 static void *read_object(const unsigned char *sha1, enum object_type *type,
                         unsigned long *size);
 
@@ -2670,8 +2676,10 @@ void *unpack_entry(struct packed_git *p, off_t obj_offset,
                free(external_base);
        }
 
-       *final_type = type;
-       *final_size = size;
+       if (final_type)
+               *final_type = type;
+       if (final_size)
+               *final_size = size;
 
        unuse_pack(&w_curs);
 
@@ -2905,6 +2913,7 @@ static int sha1_loose_object_info(const unsigned char *sha1,
        git_zstream stream;
        char hdr[32];
        struct strbuf hdrbuf = STRBUF_INIT;
+       unsigned long size_scratch;
 
        if (oi->delta_base_sha1)
                hashclr(oi->delta_base_sha1);
@@ -2917,7 +2926,7 @@ static int sha1_loose_object_info(const unsigned char *sha1,
         * return value implicitly indicates whether the
         * object even exists.
         */
-       if (!oi->typep && !oi->typename && !oi->sizep) {
+       if (!oi->typep && !oi->typename && !oi->sizep && !oi->contentp) {
                const char *path;
                struct stat st;
                if (stat_sha1_file(sha1, &st, &path) < 0)
@@ -2930,9 +2939,13 @@ static int sha1_loose_object_info(const unsigned char *sha1,
        map = map_sha1_file(sha1, &mapsize);
        if (!map)
                return -1;
+
+       if (!oi->sizep)
+               oi->sizep = &size_scratch;
+
        if (oi->disk_sizep)
                *oi->disk_sizep = mapsize;
-       if ((flags & LOOKUP_UNKNOWN_OBJECT)) {
+       if ((flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE)) {
                if (unpack_sha1_header_to_strbuf(&stream, map, mapsize, hdr, sizeof(hdr), &hdrbuf) < 0)
                        status = error("unable to unpack %s header with --allow-unknown-type",
                                       sha1_to_hex(sha1));
@@ -2947,36 +2960,52 @@ static int sha1_loose_object_info(const unsigned char *sha1,
                                       sha1_to_hex(sha1));
        } else if ((status = parse_sha1_header_extended(hdr, oi, flags)) < 0)
                status = error("unable to parse %s header", sha1_to_hex(sha1));
-       git_inflate_end(&stream);
+
+       if (status >= 0 && oi->contentp)
+               *oi->contentp = unpack_sha1_rest(&stream, hdr,
+                                                *oi->sizep, sha1);
+       else
+               git_inflate_end(&stream);
+
        munmap(map, mapsize);
        if (status && oi->typep)
                *oi->typep = status;
+       if (oi->sizep == &size_scratch)
+               oi->sizep = NULL;
        strbuf_release(&hdrbuf);
        return (status < 0) ? status : 0;
 }
 
 int sha1_object_info_extended(const unsigned char *sha1, struct object_info *oi, unsigned flags)
 {
-       struct cached_object *co;
+       static struct object_info blank_oi = OBJECT_INFO_INIT;
        struct pack_entry e;
        int rtype;
-       enum object_type real_type;
-       const unsigned char *real = lookup_replace_object_extended(sha1, flags);
-
-       co = find_cached_object(real);
-       if (co) {
-               if (oi->typep)
-                       *(oi->typep) = co->type;
-               if (oi->sizep)
-                       *(oi->sizep) = co->size;
-               if (oi->disk_sizep)
-                       *(oi->disk_sizep) = 0;
-               if (oi->delta_base_sha1)
-                       hashclr(oi->delta_base_sha1);
-               if (oi->typename)
-                       strbuf_addstr(oi->typename, typename(co->type));
-               oi->whence = OI_CACHED;
-               return 0;
+       const unsigned char *real = (flags & OBJECT_INFO_LOOKUP_REPLACE) ?
+                                   lookup_replace_object(sha1) :
+                                   sha1;
+
+       if (!oi)
+               oi = &blank_oi;
+
+       if (!(flags & OBJECT_INFO_SKIP_CACHED)) {
+               struct cached_object *co = find_cached_object(real);
+               if (co) {
+                       if (oi->typep)
+                               *(oi->typep) = co->type;
+                       if (oi->sizep)
+                               *(oi->sizep) = co->size;
+                       if (oi->disk_sizep)
+                               *(oi->disk_sizep) = 0;
+                       if (oi->delta_base_sha1)
+                               hashclr(oi->delta_base_sha1);
+                       if (oi->typename)
+                               strbuf_addstr(oi->typename, typename(co->type));
+                       if (oi->contentp)
+                               *oi->contentp = xmemdupz(co->buf, co->size);
+                       oi->whence = OI_CACHED;
+                       return 0;
+               }
        }
 
        if (!find_pack_entry(real, &e)) {
@@ -2987,23 +3016,25 @@ int sha1_object_info_extended(const unsigned char *sha1, struct object_info *oi,
                }
 
                /* Not a loose object; someone else may have just packed it. */
-               reprepare_packed_git();
-               if (!find_pack_entry(real, &e))
+               if (flags & OBJECT_INFO_QUICK) {
                        return -1;
+               } else {
+                       reprepare_packed_git();
+                       if (!find_pack_entry(real, &e))
+                               return -1;
+               }
        }
 
-       /*
-        * packed_object_info() does not follow the delta chain to
-        * find out the real type, unless it is given oi->typep.
-        */
-       if (oi->typename && !oi->typep)
-               oi->typep = &real_type;
+       if (oi == &blank_oi)
+               /*
+                * We know that the caller doesn't actually need the
+                * information below, so return early.
+                */
+               return 0;
 
        rtype = packed_object_info(e.p, e.offset, oi);
        if (rtype < 0) {
                mark_bad_packed_object(e.p, real);
-               if (oi->typep == &real_type)
-                       oi->typep = NULL;
                return sha1_object_info_extended(real, oi, 0);
        } else if (in_delta_base_cache(e.p, e.offset)) {
                oi->whence = OI_DBCACHED;
@@ -3014,10 +3045,6 @@ int sha1_object_info_extended(const unsigned char *sha1, struct object_info *oi,
                oi->u.packed.is_delta = (rtype == OBJ_REF_DELTA ||
                                         rtype == OBJ_OFS_DELTA);
        }
-       if (oi->typename)
-               strbuf_addstr(oi->typename, typename(*oi->typep));
-       if (oi->typep == &real_type)
-               oi->typep = NULL;
 
        return 0;
 }
@@ -3030,7 +3057,8 @@ int sha1_object_info(const unsigned char *sha1, unsigned long *sizep)
 
        oi.typep = &type;
        oi.sizep = sizep;
-       if (sha1_object_info_extended(sha1, &oi, LOOKUP_REPLACE_OBJECT) < 0)
+       if (sha1_object_info_extended(sha1, &oi,
+                                     OBJECT_INFO_LOOKUP_REPLACE) < 0)
                return -1;
        return type;
 }
@@ -3080,28 +3108,15 @@ int pretend_sha1_file(void *buf, unsigned long len, enum object_type type,
 static void *read_object(const unsigned char *sha1, enum object_type *type,
                         unsigned long *size)
 {
-       unsigned long mapsize;
-       void *map, *buf;
-       struct cached_object *co;
-
-       co = find_cached_object(sha1);
-       if (co) {
-               *type = co->type;
-               *size = co->size;
-               return xmemdupz(co->buf, co->size);
-       }
+       struct object_info oi = OBJECT_INFO_INIT;
+       void *content;
+       oi.typep = type;
+       oi.sizep = size;
+       oi.contentp = &content;
 
-       buf = read_packed_sha1(sha1, type, size);
-       if (buf)
-               return buf;
-       map = map_sha1_file(sha1, &mapsize);
-       if (map) {
-               buf = unpack_sha1_file(map, mapsize, type, size, sha1);
-               munmap(map, mapsize);
-               return buf;
-       }
-       reprepare_packed_git();
-       return read_packed_sha1(sha1, type, size);
+       if (sha1_object_info_extended(sha1, &oi, 0) < 0)
+               return NULL;
+       return content;
 }
 
 /*
@@ -3112,13 +3127,14 @@ static void *read_object(const unsigned char *sha1, enum object_type *type,
 void *read_sha1_file_extended(const unsigned char *sha1,
                              enum object_type *type,
                              unsigned long *size,
-                             unsigned flag)
+                             int lookup_replace)
 {
        void *data;
        const struct packed_git *p;
        const char *path;
        struct stat st;
-       const unsigned char *repl = lookup_replace_object_extended(sha1, flag);
+       const unsigned char *repl = lookup_replace ? lookup_replace_object(sha1)
+                                                  : sha1;
 
        errno = 0;
        data = read_object(repl, type, size);
@@ -3479,18 +3495,10 @@ int has_sha1_pack(const unsigned char *sha1)
 
 int has_sha1_file_with_flags(const unsigned char *sha1, int flags)
 {
-       struct pack_entry e;
-
        if (!startup_info->have_repository)
                return 0;
-       if (find_pack_entry(sha1, &e))
-               return 1;
-       if (has_loose_object(sha1))
-               return 1;
-       if (flags & HAS_SHA1_QUICK)
-               return 0;
-       reprepare_packed_git();
-       return find_pack_entry(sha1, &e);
+       return sha1_object_info_extended(sha1, NULL,
+                                        flags | OBJECT_INFO_SKIP_CACHED) >= 0;
 }
 
 int has_object_file(const struct object_id *oid)
@@ -3546,7 +3554,7 @@ static int index_mem(unsigned char *sha1, void *buf, size_t size,
         */
        if ((type == OBJ_BLOB) && path) {
                struct strbuf nbuf = STRBUF_INIT;
-               if (convert_to_git(path, buf, size, &nbuf,
+               if (convert_to_git(&the_index, path, buf, size, &nbuf,
                                   write_object ? safe_crlf : SAFE_CRLF_FALSE)) {
                        buf = strbuf_detach(&nbuf, &size);
                        re_allocated = 1;
@@ -3580,7 +3588,7 @@ static int index_stream_convert_blob(unsigned char *sha1, int fd,
        assert(path);
        assert(would_convert_to_git_filter_fd(path));
 
-       convert_to_git_filter_fd(path, fd, &sbuf,
+       convert_to_git_filter_fd(&the_index, path, fd, &sbuf,
                                 write_object ? safe_crlf : SAFE_CRLF_FALSE);
 
        if (write_object)
@@ -3668,7 +3676,7 @@ int index_fd(unsigned char *sha1, int fd, struct stat *st,
        else if (!S_ISREG(st->st_mode))
                ret = index_pipe(sha1, fd, type, path, flags);
        else if (st->st_size <= big_file_threshold || type != OBJ_BLOB ||
-                (path && would_convert_to_git(path)))
+                (path && would_convert_to_git(&the_index, path)))
                ret = index_core(sha1, fd, xsize_t(st->st_size), type, path,
                                 flags);
        else
@@ -3735,22 +3743,32 @@ void assert_sha1_type(const unsigned char *sha1, enum object_type expect)
                    typename(expect));
 }
 
-static int for_each_file_in_obj_subdir(int subdir_nr,
-                                      struct strbuf *path,
-                                      each_loose_object_fn obj_cb,
-                                      each_loose_cruft_fn cruft_cb,
-                                      each_loose_subdir_fn subdir_cb,
-                                      void *data)
+int for_each_file_in_obj_subdir(unsigned int subdir_nr,
+                               struct strbuf *path,
+                               each_loose_object_fn obj_cb,
+                               each_loose_cruft_fn cruft_cb,
+                               each_loose_subdir_fn subdir_cb,
+                               void *data)
 {
-       size_t baselen = path->len;
-       DIR *dir = opendir(path->buf);
+       size_t origlen, baselen;
+       DIR *dir;
        struct dirent *de;
        int r = 0;
 
+       if (subdir_nr > 0xff)
+               BUG("invalid loose object subdirectory: %x", subdir_nr);
+
+       origlen = path->len;
+       strbuf_complete(path, '/');
+       strbuf_addf(path, "%02x", subdir_nr);
+       baselen = path->len;
+
+       dir = opendir(path->buf);
        if (!dir) {
-               if (errno == ENOENT)
-                       return 0;
-               return error_errno("unable to open %s", path->buf);
+               if (errno != ENOENT)
+                       r = error_errno("unable to open %s", path->buf);
+               strbuf_setlen(path, origlen);
+               return r;
        }
 
        while ((de = readdir(dir))) {
@@ -3788,6 +3806,8 @@ static int for_each_file_in_obj_subdir(int subdir_nr,
        if (!r && subdir_cb)
                r = subdir_cb(subdir_nr, path->buf, data);
 
+       strbuf_setlen(path, origlen);
+
        return r;
 }
 
@@ -3797,15 +3817,12 @@ int for_each_loose_file_in_objdir_buf(struct strbuf *path,
                            each_loose_subdir_fn subdir_cb,
                            void *data)
 {
-       size_t baselen = path->len;
        int r = 0;
        int i;
 
        for (i = 0; i < 256; i++) {
-               strbuf_addf(path, "/%02x", i);
                r = for_each_file_in_obj_subdir(i, path, obj_cb, cruft_cb,
                                                subdir_cb, data);
-               strbuf_setlen(path, baselen);
                if (r)
                        break;
        }