Limit the size of the new delta_base_cache
[gitweb.git] / builtin-pack-objects.c
index 3824ee33acb2f984419ab79c2f3ecc306d3e5ea5..73d448b890d61b79793290b6b9b9aceea0a89cdc 100644 (file)
@@ -23,7 +23,7 @@ git-pack-objects [{ -q | --progress | --all-progress }] \n\
 struct object_entry {
        unsigned char sha1[20];
        unsigned long size;     /* uncompressed size */
-       unsigned long offset;   /* offset into the final pack file;
+       off_t offset;   /* offset into the final pack file;
                                 * nonzero if already written.
                                 */
        unsigned int depth;     /* delta depth */
@@ -35,7 +35,7 @@ struct object_entry {
 #define in_pack_header_size delta_size /* only when reusing pack data */
        struct object_entry *delta;     /* delta base object */
        struct packed_git *in_pack;     /* already in pack */
-       unsigned int in_pack_offset;
+       off_t in_pack_offset;
        struct object_entry *delta_child; /* deltified objects who bases me */
        struct object_entry *delta_sibling; /* other deltified objects who
                                             * uses the same base as me
@@ -68,7 +68,7 @@ static int allow_ofs_delta;
 
 static struct object_entry **sorted_by_sha, **sorted_by_type;
 static struct object_entry *objects;
-static int nr_objects, nr_alloc, nr_result;
+static uint32_t nr_objects, nr_alloc, nr_result;
 static const char *base_name;
 static unsigned char pack_file_sha1[20];
 static int progress = 1;
@@ -101,7 +101,7 @@ static int object_ix_hashsz;
  * get the object sha1 from the main index.
  */
 struct revindex_entry {
-       unsigned int offset;
+       off_t offset;
        unsigned int nr;
 };
 struct pack_revindex {
@@ -114,10 +114,8 @@ static int pack_revindex_hashsz;
 /*
  * stats
  */
-static int written;
-static int written_delta;
-static int reused;
-static int reused_delta;
+static uint32_t written, written_delta;
+static uint32_t reused, reused_delta;
 
 static int pack_revindex_ix(struct packed_git *p)
 {
@@ -168,11 +166,12 @@ static void prepare_pack_revindex(struct pack_revindex *rix)
        struct packed_git *p = rix->p;
        int num_ent = num_packed_objects(p);
        int i;
-       void *index = p->index_base + 256;
+       const char *index = p->index_data;
 
+       index += 4 * 256;
        rix->revindex = xmalloc(sizeof(*rix->revindex) * (num_ent + 1));
        for (i = 0; i < num_ent; i++) {
-               unsigned int hl = *((unsigned int *)((char *) index + 24*i));
+               uint32_t hl = *((uint32_t *)(index + 24 * i));
                rix->revindex[i].offset = ntohl(hl);
                rix->revindex[i].nr = i;
        }
@@ -185,7 +184,7 @@ static void prepare_pack_revindex(struct pack_revindex *rix)
 }
 
 static struct revindex_entry * find_packed_object(struct packed_git *p,
-                                                 unsigned int ofs)
+                                                 off_t ofs)
 {
        int num;
        int lo, hi;
@@ -213,25 +212,24 @@ static struct revindex_entry * find_packed_object(struct packed_git *p,
        die("internal error: pack revindex corrupt");
 }
 
-static unsigned long find_packed_object_size(struct packed_git *p,
-                                            unsigned long ofs)
+static off_t find_packed_object_size(struct packed_git *p, off_t ofs)
 {
        struct revindex_entry *entry = find_packed_object(p, ofs);
        return entry[1].offset - ofs;
 }
 
-static unsigned char *find_packed_object_name(struct packed_git *p,
-                                             unsigned long ofs)
+static const unsigned char *find_packed_object_name(struct packed_git *p,
+                                                   off_t ofs)
 {
        struct revindex_entry *entry = find_packed_object(p, ofs);
-       return (unsigned char *)(p->index_base + 256) + 24 * entry->nr + 4;
+       return ((unsigned char *)p->index_data) + 4 * 256 + 24 * entry->nr + 4;
 }
 
 static void *delta_against(void *buf, unsigned long size, struct object_entry *entry)
 {
        unsigned long othersize, delta_size;
-       char type[10];
-       void *otherbuf = read_sha1_file(entry->delta->sha1, type, &othersize);
+       enum object_type type;
+       void *otherbuf = read_sha1_file(entry->delta->sha1, &type, &othersize);
        void *delta_buf;
 
        if (!otherbuf)
@@ -278,8 +276,8 @@ static int encode_header(enum object_type type, unsigned long size, unsigned cha
  */
 static int check_pack_inflate(struct packed_git *p,
                struct pack_window **w_curs,
-               unsigned long offset,
-               unsigned long len,
+               off_t offset,
+               off_t len,
                unsigned long expect)
 {
        z_stream stream;
@@ -305,8 +303,8 @@ static int check_pack_inflate(struct packed_git *p,
 static void copy_pack_data(struct sha1file *f,
                struct packed_git *p,
                struct pack_window **w_curs,
-               unsigned long offset,
-               unsigned long len)
+               off_t offset,
+               off_t len)
 {
        unsigned char *in;
        unsigned int avail;
@@ -314,7 +312,7 @@ static void copy_pack_data(struct sha1file *f,
        while (len) {
                in = use_pack(p, w_curs, offset, &avail);
                if (avail > len)
-                       avail = len;
+                       avail = (unsigned int)len;
                sha1write(f, in, avail);
                offset += avail;
                len -= avail;
@@ -371,14 +369,15 @@ static int revalidate_loose_object(struct object_entry *entry,
        return check_loose_inflate(map, mapsize, size);
 }
 
-static unsigned long write_object(struct sha1file *f,
+static off_t write_object(struct sha1file *f,
                                  struct object_entry *entry)
 {
        unsigned long size;
-       char type[10];
+       enum object_type type;
        void *buf;
        unsigned char header[10];
-       unsigned hdrlen, datalen;
+       unsigned hdrlen;
+       off_t datalen;
        enum object_type obj_type;
        int to_reuse = 0;
 
@@ -416,7 +415,7 @@ static unsigned long write_object(struct sha1file *f,
        }
 
        if (!to_reuse) {
-               buf = read_sha1_file(entry->sha1, type, &size);
+               buf = read_sha1_file(entry->sha1, &type, &size);
                if (!buf)
                        die("unable to read %s", sha1_to_hex(entry->sha1));
                if (size != entry->size)
@@ -441,7 +440,7 @@ static unsigned long write_object(struct sha1file *f,
                         * encoding of the relative offset for the delta
                         * base from this object's position in the pack.
                         */
-                       unsigned long ofs = entry->offset - entry->delta->offset;
+                       off_t ofs = entry->offset - entry->delta->offset;
                        unsigned pos = sizeof(header) - 1;
                        header[pos] = ofs & 127;
                        while (ofs >>= 7)
@@ -462,7 +461,7 @@ static unsigned long write_object(struct sha1file *f,
        else {
                struct packed_git *p = entry->in_pack;
                struct pack_window *w_curs = NULL;
-               unsigned long offset;
+               off_t offset;
 
                if (entry->delta) {
                        obj_type = (allow_ofs_delta && entry->delta->offset) ?
@@ -472,7 +471,7 @@ static unsigned long write_object(struct sha1file *f,
                hdrlen = encode_header(obj_type, entry->size, header);
                sha1write(f, header, hdrlen);
                if (obj_type == OBJ_OFS_DELTA) {
-                       unsigned long ofs = entry->offset - entry->delta->offset;
+                       off_t ofs = entry->offset - entry->delta->offset;
                        unsigned pos = sizeof(header) - 1;
                        header[pos] = ofs & 127;
                        while (ofs >>= 7)
@@ -500,9 +499,9 @@ static unsigned long write_object(struct sha1file *f,
        return hdrlen + datalen;
 }
 
-static unsigned long write_one(struct sha1file *f,
+static off_t write_one(struct sha1file *f,
                               struct object_entry *e,
-                              unsigned long offset)
+                              off_t offset)
 {
        if (e->offset || e->preferred_base)
                /* offset starts from header size and cannot be zero
@@ -518,9 +517,9 @@ static unsigned long write_one(struct sha1file *f,
 
 static void write_pack_file(void)
 {
-       int i;
+       uint32_t i;
        struct sha1file *f;
-       unsigned long offset;
+       off_t offset;
        struct pack_header hdr;
        unsigned last_percent = 999;
        int do_progress = progress;
@@ -533,7 +532,7 @@ static void write_pack_file(void)
                f = sha1create("%s-%s.%s", base_name,
                               sha1_to_hex(object_list_sha1), "pack");
        if (do_progress)
-               fprintf(stderr, "Writing %d objects.\n", nr_result);
+               fprintf(stderr, "Writing %u objects.\n", nr_result);
 
        hdr.hdr_signature = htonl(PACK_SIGNATURE);
        hdr.hdr_version = htonl(PACK_VERSION);
@@ -558,13 +557,13 @@ static void write_pack_file(void)
                fputc('\n', stderr);
  done:
        if (written != nr_result)
-               die("wrote %d objects while expecting %d", written, nr_result);
+               die("wrote %u objects while expecting %u", written, nr_result);
        sha1close(f, pack_file_sha1, 1);
 }
 
 static void write_index_file(void)
 {
-       int i;
+       uint32_t i;
        struct sha1file *f = sha1create("%s-%s.%s", base_name,
                                        sha1_to_hex(object_list_sha1), "idx");
        struct object_entry **list = sorted_by_sha;
@@ -633,7 +632,7 @@ static struct object_entry *locate_object_entry(const unsigned char *sha1)
 
 static void rehash_objects(void)
 {
-       int i;
+       uint32_t i;
        struct object_entry *oe;
 
        object_ix_hashsz = nr_objects * 3;
@@ -670,16 +669,16 @@ static unsigned name_hash(const char *name)
 
 static int add_object_entry(const unsigned char *sha1, unsigned hash, int exclude)
 {
-       unsigned int idx = nr_objects;
+       uint32_t idx = nr_objects;
        struct object_entry *entry;
        struct packed_git *p;
-       unsigned int found_offset = 0;
+       off_t found_offset = 0;
        struct packed_git *found_pack = NULL;
        int ix, status = 0;
 
        if (!exclude) {
                for (p = packed_git; p; p = p->next) {
-                       unsigned long offset = find_pack_entry_one(sha1, p);
+                       off_t offset = find_pack_entry_one(sha1, p);
                        if (offset) {
                                if (incremental)
                                        return 0;
@@ -696,9 +695,8 @@ static int add_object_entry(const unsigned char *sha1, unsigned hash, int exclud
                goto already_added;
 
        if (idx >= nr_alloc) {
-               unsigned int needed = (idx + 1024) * 3 / 2;
-               objects = xrealloc(objects, needed * sizeof(*entry));
-               nr_alloc = needed;
+               nr_alloc = (idx + 1024) * 3 / 2;
+               objects = xrealloc(objects, nr_alloc * sizeof(*entry));
        }
        entry = objects + idx;
        nr_objects = idx + 1;
@@ -718,7 +716,7 @@ static int add_object_entry(const unsigned char *sha1, unsigned hash, int exclud
 
  already_added:
        if (progress_update) {
-               fprintf(stderr, "Counting objects...%d\r", nr_objects);
+               fprintf(stderr, "Counting objects...%u\r", nr_objects);
                progress_update = 0;
        }
        if (exclude)
@@ -765,7 +763,7 @@ static struct pbase_tree_cache *pbase_tree_get(const unsigned char *sha1)
        struct pbase_tree_cache *ent, *nent;
        void *data;
        unsigned long size;
-       char type[20];
+       enum object_type type;
        int neigh;
        int my_ix = pbase_tree_cache_ix(sha1);
        int available_ix = -1;
@@ -792,10 +790,10 @@ static struct pbase_tree_cache *pbase_tree_get(const unsigned char *sha1)
        /* Did not find one.  Either we got a bogus request or
         * we need to read and perhaps cache.
         */
-       data = read_sha1_file(sha1, type, &size);
+       data = read_sha1_file(sha1, &type, &size);
        if (!data)
                return NULL;
-       if (strcmp(type, tree_type)) {
+       if (type != OBJ_TREE) {
                free(data);
                return NULL;
        }
@@ -854,19 +852,19 @@ static void add_pbase_object(struct tree_desc *tree,
 
        while (tree_entry(tree,&entry)) {
                unsigned long size;
-               char type[20];
+               enum object_type type;
 
                if (entry.pathlen != cmplen ||
                    memcmp(entry.path, name, cmplen) ||
                    !has_sha1_file(entry.sha1) ||
-                   sha1_object_info(entry.sha1, type, &size))
+                   (type = sha1_object_info(entry.sha1, &size)) < 0)
                        continue;
                if (name[cmplen] != '/') {
                        unsigned hash = name_hash(fullname);
                        add_object_entry(entry.sha1, hash, 1);
                        return;
                }
-               if (!strcmp(type, tree_type)) {
+               if (type == OBJ_TREE) {
                        struct tree_desc sub;
                        struct pbase_tree_cache *tree;
                        const char *down = name+cmplen+1;
@@ -978,22 +976,20 @@ static void add_preferred_base(unsigned char *sha1)
 
 static void check_object(struct object_entry *entry)
 {
-       char type[20];
-
        if (entry->in_pack && !entry->preferred_base) {
                struct packed_git *p = entry->in_pack;
                struct pack_window *w_curs = NULL;
-               unsigned long left = p->pack_size - entry->in_pack_offset;
                unsigned long size, used;
+               unsigned int avail;
                unsigned char *buf;
                struct object_entry *base_entry = NULL;
 
-               buf = use_pack(p, &w_curs, entry->in_pack_offset, NULL);
+               buf = use_pack(p, &w_curs, entry->in_pack_offset, &avail);
 
                /* We want in_pack_type even if we do not reuse delta.
                 * There is no point not reusing non-delta representations.
                 */
-               used = unpack_object_header_gently(buf, left,
+               used = unpack_object_header_gently(buf, avail,
                                                   &entry->in_pack_type, &size);
 
                /* Check if it is delta, and the base is also an object
@@ -1001,8 +997,9 @@ static void check_object(struct object_entry *entry)
                 * delta.
                 */
                if (!no_reuse_delta) {
-                       unsigned char c, *base_name;
-                       unsigned long ofs;
+                       unsigned char c;
+                       const unsigned char *base_name;
+                       off_t ofs;
                        unsigned long used_0;
                        /* there is at least 20 bytes left in the pack */
                        switch (entry->in_pack_type) {
@@ -1062,21 +1059,10 @@ static void check_object(struct object_entry *entry)
                /* Otherwise we would do the usual */
        }
 
-       if (sha1_object_info(entry->sha1, type, &entry->size))
+       entry->type = sha1_object_info(entry->sha1, &entry->size);
+       if (entry->type < 0)
                die("unable to get type of object %s",
                    sha1_to_hex(entry->sha1));
-
-       if (!strcmp(type, commit_type)) {
-               entry->type = OBJ_COMMIT;
-       } else if (!strcmp(type, tree_type)) {
-               entry->type = OBJ_TREE;
-       } else if (!strcmp(type, blob_type)) {
-               entry->type = OBJ_BLOB;
-       } else if (!strcmp(type, tag_type)) {
-               entry->type = OBJ_TAG;
-       } else
-               die("unable to pack object %s of type %s",
-                   sha1_to_hex(entry->sha1), type);
 }
 
 static unsigned int check_delta_limit(struct object_entry *me, unsigned int n)
@@ -1094,7 +1080,7 @@ static unsigned int check_delta_limit(struct object_entry *me, unsigned int n)
 
 static void get_object_details(void)
 {
-       int i;
+       uint32_t i;
        struct object_entry *entry;
 
        prepare_pack_ix();
@@ -1133,7 +1119,7 @@ static int sort_comparator(const void *_a, const void *_b)
 static struct object_entry **create_sorted_list(entry_sort_t sort)
 {
        struct object_entry **list = xmalloc(nr_objects * sizeof(struct object_entry *));
-       int i;
+       uint32_t i;
 
        for (i = 0; i < nr_objects; i++)
                list[i] = objects + i;
@@ -1150,7 +1136,7 @@ static int sha1_sort(const struct object_entry *a, const struct object_entry *b)
 static struct object_entry **create_final_object_list(void)
 {
        struct object_entry **list;
-       int i, j;
+       uint32_t i, j;
 
        for (i = nr_result = 0; i < nr_objects; i++)
                if (!objects[i].preferred_base)
@@ -1206,7 +1192,7 @@ static int try_delta(struct unpacked *trg, struct unpacked *src,
        struct object_entry *trg_entry = trg->entry;
        struct object_entry *src_entry = src->entry;
        unsigned long trg_size, src_size, delta_size, sizediff, max_size, sz;
-       char type[10];
+       enum object_type type;
        void *delta_buf;
 
        /* Don't bother doing diffs between different types */
@@ -1257,13 +1243,13 @@ static int try_delta(struct unpacked *trg, struct unpacked *src,
 
        /* Load data if not already done */
        if (!trg->data) {
-               trg->data = read_sha1_file(trg_entry->sha1, type, &sz);
+               trg->data = read_sha1_file(trg_entry->sha1, &type, &sz);
                if (sz != trg_size)
                        die("object %s inconsistent object length (%lu vs %lu)",
                            sha1_to_hex(trg_entry->sha1), sz, trg_size);
        }
        if (!src->data) {
-               src->data = read_sha1_file(src_entry->sha1, type, &sz);
+               src->data = read_sha1_file(src_entry->sha1, &type, &sz);
                if (sz != src_size)
                        die("object %s inconsistent object length (%lu vs %lu)",
                            sha1_to_hex(src_entry->sha1), sz, src_size);
@@ -1292,20 +1278,20 @@ static void progress_interval(int signum)
 
 static void find_deltas(struct object_entry **list, int window, int depth)
 {
-       int i, idx;
+       uint32_t i = nr_objects, idx = 0, processed = 0;
        unsigned int array_size = window * sizeof(struct unpacked);
-       struct unpacked *array = xmalloc(array_size);
-       unsigned processed = 0;
+       struct unpacked *array;
        unsigned last_percent = 999;
 
+       if (!nr_objects)
+               return;
+       array = xmalloc(array_size);
        memset(array, 0, array_size);
-       i = nr_objects;
-       idx = 0;
        if (progress)
-               fprintf(stderr, "Deltifying %d objects.\n", nr_result);
+               fprintf(stderr, "Deltifying %u objects.\n", nr_result);
 
-       while (--i >= 0) {
-               struct object_entry *entry = list[i];
+       do {
+               struct object_entry *entry = list[--i];
                struct unpacked *n = array + idx;
                int j;
 
@@ -1338,7 +1324,7 @@ static void find_deltas(struct object_entry **list, int window, int depth)
 
                j = window;
                while (--j > 0) {
-                       unsigned int other_idx = idx + j;
+                       uint32_t other_idx = idx + j;
                        struct unpacked *m;
                        if (other_idx >= window)
                                other_idx -= window;
@@ -1358,7 +1344,7 @@ static void find_deltas(struct object_entry **list, int window, int depth)
                idx++;
                if (idx >= window)
                        idx = 0;
-       }
+       } while (i > 0);
 
        if (progress)
                fputc('\n', stderr);
@@ -1399,7 +1385,7 @@ static int reuse_cached_pack(unsigned char *sha1)
        }
 
        if (progress)
-               fprintf(stderr, "Reusing %d objects pack %s\n", nr_objects,
+               fprintf(stderr, "Reusing %u objects pack %s\n", nr_objects,
                        sha1_to_hex(sha1));
 
        if (pack_to_stdout) {
@@ -1550,10 +1536,13 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
        struct object_entry **list;
        int use_internal_rev_list = 0;
        int thin = 0;
-       int i;
-       const char *rp_av[64];
+       uint32_t i;
+       const char **rp_av;
+       int rp_ac_alloc = 64;
        int rp_ac;
 
+       rp_av = xcalloc(rp_ac_alloc, sizeof(*rp_av));
+
        rp_av[0] = "pack-objects";
        rp_av[1] = "--objects"; /* --thin will make it --objects-edge */
        rp_ac = 2;
@@ -1579,14 +1568,14 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
                        incremental = 1;
                        continue;
                }
-               if (!strncmp("--window=", arg, 9)) {
+               if (!prefixcmp(arg, "--window=")) {
                        char *end;
                        window = strtoul(arg+9, &end, 0);
                        if (!arg[9] || *end)
                                usage(pack_usage);
                        continue;
                }
-               if (!strncmp("--depth=", arg, 8)) {
+               if (!prefixcmp(arg, "--depth=")) {
                        char *end;
                        depth = strtoul(arg+8, &end, 0);
                        if (!arg[8] || *end)
@@ -1622,12 +1611,15 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
                        continue;
                }
                if (!strcmp("--unpacked", arg) ||
-                   !strncmp("--unpacked=", arg, 11) ||
+                   !prefixcmp(arg, "--unpacked=") ||
                    !strcmp("--reflog", arg) ||
                    !strcmp("--all", arg)) {
                        use_internal_rev_list = 1;
-                       if (ARRAY_SIZE(rp_av) - 1 <= rp_ac)
-                               die("too many internal rev-list options");
+                       if (rp_ac >= rp_ac_alloc - 1) {
+                               rp_ac_alloc = alloc_nr(rp_ac_alloc);
+                               rp_av = xrealloc(rp_av,
+                                                rp_ac_alloc * sizeof(*rp_av));
+                       }
                        rp_av[rp_ac++] = arg;
                        continue;
                }
@@ -1677,7 +1669,7 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
        }
 
        if (progress)
-               fprintf(stderr, "Done counting %d objects.\n", nr_objects);
+               fprintf(stderr, "Done counting %u objects.\n", nr_objects);
        sorted_by_sha = create_final_object_list();
        if (non_empty && !nr_result)
                return 0;
@@ -1690,7 +1682,7 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
        }
        SHA1_Final(object_list_sha1, &ctx);
        if (progress && (nr_objects != nr_result))
-               fprintf(stderr, "Result has %d objects.\n", nr_result);
+               fprintf(stderr, "Result has %u objects.\n", nr_result);
 
        if (reuse_cached_pack(object_list_sha1))
                ;
@@ -1711,7 +1703,7 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
                }
        }
        if (progress)
-               fprintf(stderr, "Total %d (delta %d), reused %d (delta %d)\n",
+               fprintf(stderr, "Total %u (delta %u), reused %u (delta %u)\n",
                        written, written_delta, reused, reused_delta);
        return 0;
 }