p5551: add a script to test fetch pack-dir rescans
[gitweb.git] / builtin / index-pack.c
index 1d2ea583a45507935ec2007d6c44f5e968a1aed5..26828c1d82bc9a997041cb464d4a3fde38e2d19d 100644 (file)
@@ -1,4 +1,5 @@
 #include "builtin.h"
+#include "config.h"
 #include "delta.h"
 #include "pack.h"
 #include "csum-file.h"
@@ -87,6 +88,7 @@ static struct progress *progress;
 static unsigned char input_buffer[4096];
 static unsigned int input_offset, input_len;
 static off_t consumed_bytes;
+static off_t max_input_size;
 static unsigned deepest_delta;
 static git_SHA_CTX input_ctx;
 static uint32_t input_crc32;
@@ -297,6 +299,8 @@ static void use(int bytes)
        if (signed_add_overflows(consumed_bytes, bytes))
                die(_("pack too large for current definition of off_t"));
        consumed_bytes += bytes;
+       if (max_input_size && consumed_bytes > max_input_size)
+               die(_("pack exceeds maximum allowed size"));
 }
 
 static const char *open_pack_file(const char *pack_name)
@@ -304,14 +308,15 @@ static const char *open_pack_file(const char *pack_name)
        if (from_stdin) {
                input_fd = 0;
                if (!pack_name) {
-                       static char tmp_file[PATH_MAX];
-                       output_fd = odb_mkstemp(tmp_file, sizeof(tmp_file),
+                       struct strbuf tmp_file = STRBUF_INIT;
+                       output_fd = odb_mkstemp(&tmp_file,
                                                "pack/tmp_pack_XXXXXX");
-                       pack_name = xstrdup(tmp_file);
-               } else
+                       pack_name = strbuf_detach(&tmp_file, NULL);
+               } else {
                        output_fd = open(pack_name, O_CREAT|O_EXCL|O_RDWR, 0600);
-               if (output_fd < 0)
-                       die_errno(_("unable to create '%s'"), pack_name);
+                       if (output_fd < 0)
+                               die_errno(_("unable to create '%s'"), pack_name);
+               }
                nothread_data.pack_fd = output_fd;
        } else {
                input_fd = open(pack_name, O_RDONLY);
@@ -384,8 +389,7 @@ static struct base_data *alloc_base_data(void)
 static void free_base_data(struct base_data *c)
 {
        if (c->data) {
-               free(c->data);
-               c->data = NULL;
+               FREE_AND_NULL(c->data);
                get_thread_data()->base_cache_used -= c->size;
        }
 }
@@ -601,8 +605,7 @@ static void *unpack_data(struct object_entry *obj,
        git_inflate_end(&stream);
        free(inbuf);
        if (consume) {
-               free(data);
-               data = NULL;
+               FREE_AND_NULL(data);
        }
        return data;
 }
@@ -743,13 +746,13 @@ static int compare_objects(const unsigned char *buf, unsigned long size,
                ssize_t len = read_istream(data->st, data->buf, size);
                if (len == 0)
                        die(_("SHA1 COLLISION FOUND WITH %s !"),
-                           sha1_to_hex(data->entry->idx.sha1));
+                           oid_to_hex(&data->entry->idx.oid));
                if (len < 0)
                        die(_("unable to read %s"),
-                           sha1_to_hex(data->entry->idx.sha1));
+                           oid_to_hex(&data->entry->idx.oid));
                if (memcmp(buf, data->buf, len))
                        die(_("SHA1 COLLISION FOUND WITH %s !"),
-                           sha1_to_hex(data->entry->idx.sha1));
+                           oid_to_hex(&data->entry->idx.oid));
                size -= len;
                buf += len;
        }
@@ -767,12 +770,12 @@ static int check_collison(struct object_entry *entry)
 
        memset(&data, 0, sizeof(data));
        data.entry = entry;
-       data.st = open_istream(entry->idx.sha1, &type, &size, NULL);
+       data.st = open_istream(entry->idx.oid.hash, &type, &size, NULL);
        if (!data.st)
                return -1;
        if (size != entry->size || type != entry->type)
                die(_("SHA1 COLLISION FOUND WITH %s !"),
-                   sha1_to_hex(entry->idx.sha1));
+                   oid_to_hex(&entry->idx.oid));
        unpack_data(entry, compare_objects, &data);
        close_istream(data.st);
        free(data.buf);
@@ -781,16 +784,19 @@ static int check_collison(struct object_entry *entry)
 
 static void sha1_object(const void *data, struct object_entry *obj_entry,
                        unsigned long size, enum object_type type,
-                       const unsigned char *sha1)
+                       const struct object_id *oid)
 {
        void *new_data = NULL;
-       int collision_test_needed;
+       int collision_test_needed = 0;
 
        assert(data || obj_entry);
 
-       read_lock();
-       collision_test_needed = has_sha1_file_with_flags(sha1, HAS_SHA1_QUICK);
-       read_unlock();
+       if (startup_info->have_repository) {
+               read_lock();
+               collision_test_needed =
+                       has_sha1_file_with_flags(oid->hash, OBJECT_INFO_QUICK);
+               read_unlock();
+       }
 
        if (collision_test_needed && !data) {
                read_lock();
@@ -803,29 +809,31 @@ static void sha1_object(const void *data, struct object_entry *obj_entry,
                enum object_type has_type;
                unsigned long has_size;
                read_lock();
-               has_type = sha1_object_info(sha1, &has_size);
+               has_type = sha1_object_info(oid->hash, &has_size);
+               if (has_type < 0)
+                       die(_("cannot read existing object info %s"), oid_to_hex(oid));
                if (has_type != type || has_size != size)
-                       die(_("SHA1 COLLISION FOUND WITH %s !"), sha1_to_hex(sha1));
-               has_data = read_sha1_file(sha1, &has_type, &has_size);
+                       die(_("SHA1 COLLISION FOUND WITH %s !"), oid_to_hex(oid));
+               has_data = read_sha1_file(oid->hash, &has_type, &has_size);
                read_unlock();
                if (!data)
                        data = new_data = get_data_from_pack(obj_entry);
                if (!has_data)
-                       die(_("cannot read existing object %s"), sha1_to_hex(sha1));
+                       die(_("cannot read existing object %s"), oid_to_hex(oid));
                if (size != has_size || type != has_type ||
                    memcmp(data, has_data, size) != 0)
-                       die(_("SHA1 COLLISION FOUND WITH %s !"), sha1_to_hex(sha1));
+                       die(_("SHA1 COLLISION FOUND WITH %s !"), oid_to_hex(oid));
                free(has_data);
        }
 
        if (strict) {
                read_lock();
                if (type == OBJ_BLOB) {
-                       struct blob *blob = lookup_blob(sha1);
+                       struct blob *blob = lookup_blob(oid);
                        if (blob)
                                blob->object.flags |= FLAG_CHECKED;
                        else
-                               die(_("invalid blob object %s"), sha1_to_hex(sha1));
+                               die(_("invalid blob object %s"), oid_to_hex(oid));
                } else {
                        struct object *obj;
                        int eaten;
@@ -837,7 +845,8 @@ static void sha1_object(const void *data, struct object_entry *obj_entry,
                         * we do not need to free the memory here, as the
                         * buf is deleted by the caller.
                         */
-                       obj = parse_object_buffer(sha1, type, size, buf, &eaten);
+                       obj = parse_object_buffer(oid, type, size, buf,
+                                                 &eaten);
                        if (!obj)
                                die(_("invalid %s"), typename(type));
                        if (do_fsck_object &&
@@ -949,9 +958,10 @@ static void resolve_delta(struct object_entry *delta_obj,
        if (!result->data)
                bad_object(delta_obj->idx.offset, _("failed to apply delta"));
        hash_sha1_file(result->data, result->size,
-                      typename(delta_obj->real_type), delta_obj->idx.sha1);
+                      typename(delta_obj->real_type),
+                      delta_obj->idx.oid.hash);
        sha1_object(result->data, NULL, result->size, delta_obj->real_type,
-                   delta_obj->idx.sha1);
+                   &delta_obj->idx.oid);
        counter_lock();
        nr_resolved_deltas++;
        counter_unlock();
@@ -981,7 +991,7 @@ static struct base_data *find_unresolved_deltas_1(struct base_data *base,
                                                  struct base_data *prev_base)
 {
        if (base->ref_last == -1 && base->ofs_last == -1) {
-               find_ref_delta_children(base->obj->idx.sha1,
+               find_ref_delta_children(base->obj->idx.oid.hash,
                                        &base->ref_first, &base->ref_last,
                                        OBJ_REF_DELTA);
 
@@ -1122,7 +1132,8 @@ static void parse_pack_objects(unsigned char *sha1)
        for (i = 0; i < nr_objects; i++) {
                struct object_entry *obj = &objects[i];
                void *data = unpack_raw_entry(obj, &ofs_delta->offset,
-                                             ref_delta_sha1, obj->idx.sha1);
+                                             ref_delta_sha1,
+                                             obj->idx.oid.hash);
                obj->real_type = obj->type;
                if (obj->type == OBJ_OFS_DELTA) {
                        nr_ofs_deltas++;
@@ -1138,7 +1149,8 @@ static void parse_pack_objects(unsigned char *sha1)
                        obj->real_type = OBJ_BAD;
                        nr_delays++;
                } else
-                       sha1_object(data, NULL, obj->size, obj->type, obj->idx.sha1);
+                       sha1_object(data, NULL, obj->size, obj->type,
+                                   &obj->idx.oid);
                free(data);
                display_progress(progress, i+1);
        }
@@ -1164,7 +1176,8 @@ static void parse_pack_objects(unsigned char *sha1)
                if (obj->real_type != OBJ_BAD)
                        continue;
                obj->real_type = obj->type;
-               sha1_object(NULL, obj, obj->size, obj->type, obj->idx.sha1);
+               sha1_object(NULL, obj, obj->size, obj->type,
+                           &obj->idx.oid);
                nr_delays--;
        }
        if (nr_delays)
@@ -1187,10 +1200,8 @@ static void resolve_deltas(void)
                return;
 
        /* Sort deltas by base SHA1/offset for fast searching */
-       qsort(ofs_deltas, nr_ofs_deltas, sizeof(struct ofs_delta_entry),
-             compare_ofs_delta_entry);
-       qsort(ref_deltas, nr_ref_deltas, sizeof(struct ref_delta_entry),
-             compare_ref_delta_entry);
+       QSORT(ofs_deltas, nr_ofs_deltas, compare_ofs_delta_entry);
+       QSORT(ref_deltas, nr_ref_deltas, compare_ref_delta_entry);
 
        if (verbose || show_resolving_progress)
                progress = start_progress(_("Resolving deltas"),
@@ -1324,7 +1335,7 @@ static struct object_entry *append_obj_to_pack(struct sha1file *f,
        obj[1].idx.offset += write_compressed(f, buf, size);
        obj[0].idx.crc32 = crc32_end(f);
        sha1flush(f);
-       hashcpy(obj->idx.sha1, sha1);
+       hashcpy(obj->idx.oid.hash, sha1);
        return obj;
 }
 
@@ -1353,7 +1364,7 @@ static void fix_unresolved_deltas(struct sha1file *f)
        ALLOC_ARRAY(sorted_by_pos, nr_ref_deltas);
        for (i = 0; i < nr_ref_deltas; i++)
                sorted_by_pos[i] = &ref_deltas[i];
-       qsort(sorted_by_pos, nr_ref_deltas, sizeof(*sorted_by_pos), delta_pos_compare);
+       QSORT(sorted_by_pos, nr_ref_deltas, delta_pos_compare);
 
        for (i = 0; i < nr_ref_deltas; i++) {
                struct ref_delta_entry *d = sorted_by_pos[i];
@@ -1383,7 +1394,9 @@ static void final(const char *final_pack_name, const char *curr_pack_name,
                  unsigned char *sha1)
 {
        const char *report = "pack";
-       char name[PATH_MAX];
+       struct strbuf pack_name = STRBUF_INIT;
+       struct strbuf index_name = STRBUF_INIT;
+       struct strbuf keep_name_buf = STRBUF_INIT;
        int err;
 
        if (!from_stdin) {
@@ -1399,14 +1412,13 @@ static void final(const char *final_pack_name, const char *curr_pack_name,
                int keep_fd, keep_msg_len = strlen(keep_msg);
 
                if (!keep_name)
-                       keep_fd = odb_pack_keep(name, sizeof(name), sha1);
-               else
-                       keep_fd = open(keep_name, O_RDWR|O_CREAT|O_EXCL, 0600);
+                       keep_name = odb_pack_name(&keep_name_buf, sha1, "keep");
 
+               keep_fd = odb_pack_keep(keep_name);
                if (keep_fd < 0) {
                        if (errno != EEXIST)
                                die_errno(_("cannot write keep file '%s'"),
-                                         keep_name ? keep_name : name);
+                                         keep_name);
                } else {
                        if (keep_msg_len > 0) {
                                write_or_die(keep_fd, keep_msg, keep_msg_len);
@@ -1414,28 +1426,22 @@ static void final(const char *final_pack_name, const char *curr_pack_name,
                        }
                        if (close(keep_fd) != 0)
                                die_errno(_("cannot close written keep file '%s'"),
-                                         keep_name ? keep_name : name);
+                                         keep_name);
                        report = "keep";
                }
        }
 
        if (final_pack_name != curr_pack_name) {
-               if (!final_pack_name) {
-                       snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
-                                get_object_directory(), sha1_to_hex(sha1));
-                       final_pack_name = name;
-               }
+               if (!final_pack_name)
+                       final_pack_name = odb_pack_name(&pack_name, sha1, "pack");
                if (finalize_object_file(curr_pack_name, final_pack_name))
                        die(_("cannot store pack file"));
        } else if (from_stdin)
                chmod(final_pack_name, 0444);
 
        if (final_index_name != curr_index_name) {
-               if (!final_index_name) {
-                       snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
-                                get_object_directory(), sha1_to_hex(sha1));
-                       final_index_name = name;
-               }
+               if (!final_index_name)
+                       final_index_name = odb_pack_name(&index_name, sha1, "idx");
                if (finalize_object_file(curr_index_name, final_index_name))
                        die(_("cannot store index file"));
        } else
@@ -1444,10 +1450,11 @@ static void final(const char *final_pack_name, const char *curr_pack_name,
        if (!from_stdin) {
                printf("%s\n", sha1_to_hex(sha1));
        } else {
-               char buf[48];
-               int len = snprintf(buf, sizeof(buf), "%s\t%s\n",
-                                  report, sha1_to_hex(sha1));
-               write_or_die(1, buf, len);
+               struct strbuf buf = STRBUF_INIT;
+
+               strbuf_addf(&buf, "%s\t%s\n", report, sha1_to_hex(sha1));
+               write_or_die(1, buf.buf, buf.len);
+               strbuf_release(&buf);
 
                /*
                 * Let's just mimic git-unpack-objects here and write
@@ -1461,6 +1468,10 @@ static void final(const char *final_pack_name, const char *curr_pack_name,
                        input_offset += err;
                }
        }
+
+       strbuf_release(&index_name);
+       strbuf_release(&pack_name);
+       strbuf_release(&keep_name_buf);
 }
 
 static int git_index_pack_config(const char *k, const char *v, void *cb)
@@ -1530,8 +1541,7 @@ static void read_v2_anomalous_offsets(struct packed_git *p,
                opts->anomaly[opts->anomaly_nr++] = ntohl(idx2[off * 2 + 1]);
        }
 
-       if (1 < opts->anomaly_nr)
-               qsort(opts->anomaly, opts->anomaly_nr, sizeof(uint32_t), cmp_uint32);
+       QSORT(opts->anomaly, opts->anomaly_nr, cmp_uint32);
 }
 
 static void read_idx_option(struct pack_idx_option *opts, const char *pack_name)
@@ -1576,13 +1586,14 @@ static void show_pack_info(int stat_only)
                if (stat_only)
                        continue;
                printf("%s %-6s %lu %lu %"PRIuMAX,
-                      sha1_to_hex(obj->idx.sha1),
+                      oid_to_hex(&obj->idx.oid),
                       typename(obj->real_type), obj->size,
                       (unsigned long)(obj[1].idx.offset - obj->idx.offset),
                       (uintmax_t)obj->idx.offset);
                if (is_delta_type(obj->type)) {
                        struct object_entry *bobj = &objects[obj_stat[i].base_object_no];
-                       printf(" %u %s", obj_stat[i].delta_depth, sha1_to_hex(bobj->idx.sha1));
+                       printf(" %u %s", obj_stat[i].delta_depth,
+                              oid_to_hex(&bobj->idx.oid));
                }
                putchar('\n');
        }
@@ -1714,6 +1725,8 @@ int cmd_index_pack(int argc, const char **argv, const char *prefix)
                                        opts.off32_limit = strtoul(c+1, &c, 0);
                                if (*c || opts.off32_limit & 0x80000000)
                                        die(_("bad %s"), arg);
+                       } else if (skip_prefix(arg, "--max-input-size=", &arg)) {
+                               max_input_size = strtoumax(arg, NULL, 10);
                        } else
                                usage(index_pack_usage);
                        continue;
@@ -1728,6 +1741,8 @@ int cmd_index_pack(int argc, const char **argv, const char *prefix)
                usage(index_pack_usage);
        if (fix_thin_pack && !from_stdin)
                die(_("--fix-thin cannot be used without --stdin"));
+       if (from_stdin && !startup_info->have_repository)
+               die(_("--stdin requires a git repository"));
        if (!index_name && pack_name)
                index_name = derive_filename(pack_name, ".idx", &index_name_buf);
        if (keep_msg && !keep_name && pack_name)