Merge branch 'rs/fsck-null-return-from-lookup'
[gitweb.git] / builtin / fsck.c
index f01b81eebfebc1c221e81c44e3f14a4e2781f0a7..56afe405b8072b3099a23661ac586f673e54ff0e 100644 (file)
@@ -1,5 +1,6 @@
 #include "builtin.h"
 #include "cache.h"
+#include "config.h"
 #include "commit.h"
 #include "tree.h"
 #include "blob.h"
 #include "progress.h"
 #include "streaming.h"
 #include "decorate.h"
+#include "packfile.h"
 
 #define REACHABLE 0x0001
 #define SEEN      0x0002
 #define HAS_OBJ   0x0004
+/* This flag is set if something points to this object. */
+#define USED      0x0008
 
 static int show_root;
 static int show_tags;
@@ -56,6 +60,23 @@ static const char *describe_object(struct object *obj)
        return buf.buf;
 }
 
+static const char *printable_type(struct object *obj)
+{
+       const char *ret;
+
+       if (obj->type == OBJ_NONE) {
+               enum object_type type = sha1_object_info(obj->oid.hash, NULL);
+               if (type > 0)
+                       object_as_type(obj, type, 0);
+       }
+
+       ret = typename(obj->type);
+       if (!ret)
+               ret = "unknown";
+
+       return ret;
+}
+
 static int fsck_config(const char *var, const char *value, void *cb)
 {
        if (strcmp(var, "fsck.skiplist") == 0) {
@@ -83,7 +104,7 @@ static void objreport(struct object *obj, const char *msg_type,
                        const char *err)
 {
        fprintf(stderr, "%s in %s %s: %s\n",
-               msg_type, typename(obj->type), describe_object(obj), err);
+               msg_type, printable_type(obj), describe_object(obj), err);
 }
 
 static int objerror(struct object *obj, const char *err)
@@ -114,7 +135,7 @@ static int mark_object(struct object *obj, int type, void *data, struct fsck_opt
        if (!obj) {
                /* ... these references to parent->fld are safe here */
                printf("broken link from %7s %s\n",
-                          typename(parent->type), describe_object(parent));
+                          printable_type(parent), describe_object(parent));
                printf("broken link from %7s %s\n",
                           (type == OBJ_ANY ? "unknown" : typename(type)), "unknown");
                errors_found |= ERROR_REACHABLE;
@@ -131,9 +152,9 @@ static int mark_object(struct object *obj, int type, void *data, struct fsck_opt
        if (!(obj->flags & HAS_OBJ)) {
                if (parent && !has_object_file(&obj->oid)) {
                        printf("broken link from %7s %s\n",
-                                typename(parent->type), describe_object(parent));
+                                printable_type(parent), describe_object(parent));
                        printf("              to %7s %s\n",
-                                typename(obj->type), describe_object(obj));
+                                printable_type(obj), describe_object(obj));
                        errors_found |= ERROR_REACHABLE;
                }
                return 1;
@@ -150,18 +171,7 @@ static void mark_object_reachable(struct object *obj)
 
 static int traverse_one_object(struct object *obj)
 {
-       int result;
-       struct tree *tree = NULL;
-
-       if (obj->type == OBJ_TREE) {
-               tree = (struct tree *)obj;
-               if (parse_tree(tree) < 0)
-                       return 1; /* error already displayed */
-       }
-       result = fsck_walk(obj, obj, &fsck_walk_options);
-       if (tree)
-               free_tree_buffer(tree);
-       return result;
+       return fsck_walk(obj, obj, &fsck_walk_options);
 }
 
 static int traverse_reachable(void)
@@ -170,14 +180,9 @@ static int traverse_reachable(void)
        unsigned int nr = 0;
        int result = 0;
        if (show_progress)
-               progress = start_progress_delay(_("Checking connectivity"), 0, 0, 2);
+               progress = start_delayed_progress(_("Checking connectivity"), 0);
        while (pending.nr) {
-               struct object_array_entry *entry;
-               struct object *obj;
-
-               entry = pending.objects + --pending.nr;
-               obj = entry->item;
-               result |= traverse_one_object(obj);
+               result |= traverse_one_object(object_array_pop(&pending));
                display_progress(progress, ++nr);
        }
        stop_progress(&progress);
@@ -188,7 +193,7 @@ static int mark_used(struct object *obj, int type, void *data, struct fsck_optio
 {
        if (!obj)
                return 1;
-       obj->used = 1;
+       obj->flags |= USED;
        return 0;
 }
 
@@ -205,9 +210,7 @@ static void check_reachable_object(struct object *obj)
        if (!(obj->flags & HAS_OBJ)) {
                if (has_sha1_pack(obj->oid.hash))
                        return; /* it is in pack - forget about it */
-               if (connectivity_only && has_object_file(&obj->oid))
-                       return;
-               printf("missing %s %s\n", typename(obj->type),
+               printf("missing %s %s\n", printable_type(obj),
                        describe_object(obj));
                errors_found |= ERROR_REACHABLE;
                return;
@@ -225,7 +228,7 @@ static void check_unreachable_object(struct object *obj)
         * to complain about it being unreachable (since it does
         * not exist).
         */
-       if (!obj->parsed)
+       if (!(obj->flags & HAS_OBJ))
                return;
 
        /*
@@ -233,13 +236,13 @@ static void check_unreachable_object(struct object *obj)
         * since this is something that is prunable.
         */
        if (show_unreachable) {
-               printf("unreachable %s %s\n", typename(obj->type),
+               printf("unreachable %s %s\n", printable_type(obj),
                        describe_object(obj));
                return;
        }
 
        /*
-        * "!used" means that nothing at all points to it, including
+        * "!USED" means that nothing at all points to it, including
         * other unreachable objects. In other words, it's the "tip"
         * of some set of unreachable objects, usually a commit that
         * got dropped.
@@ -250,9 +253,9 @@ static void check_unreachable_object(struct object *obj)
         * deleted a branch by mistake, this is a prime candidate to
         * start looking at, for example.
         */
-       if (!obj->used) {
+       if (!(obj->flags & USED)) {
                if (show_dangling)
-                       printf("dangling %s %s\n", typename(obj->type),
+                       printf("dangling %s %s\n", printable_type(obj),
                               describe_object(obj));
                if (write_lost_and_found) {
                        char *filename = git_pathdup("lost-found/%s/%s",
@@ -265,8 +268,7 @@ static void check_unreachable_object(struct object *obj)
                                free(filename);
                                return;
                        }
-                       if (!(f = fopen(filename, "w")))
-                               die_errno("Could not open '%s'", filename);
+                       f = xfopen(filename, "w");
                        if (obj->type == OBJ_BLOB) {
                                if (stream_blob_to_fd(fileno(f), &obj->oid, NULL, 1))
                                        die_errno("Could not write '%s'", filename);
@@ -320,30 +322,25 @@ static void check_connectivity(void)
 
 static int fsck_obj(struct object *obj)
 {
+       int err;
+
        if (obj->flags & SEEN)
                return 0;
        obj->flags |= SEEN;
 
        if (verbose)
                fprintf(stderr, "Checking %s %s\n",
-                       typename(obj->type), describe_object(obj));
+                       printable_type(obj), describe_object(obj));
 
        if (fsck_walk(obj, NULL, &fsck_obj_options))
                objerror(obj, "broken links");
-       if (fsck_object(obj, NULL, 0, &fsck_obj_options))
-               return -1;
-
-       if (obj->type == OBJ_TREE) {
-               struct tree *item = (struct tree *) obj;
-
-               free_tree_buffer(item);
-       }
+       err = fsck_object(obj, NULL, 0, &fsck_obj_options);
+       if (err)
+               goto out;
 
        if (obj->type == OBJ_COMMIT) {
                struct commit *commit = (struct commit *) obj;
 
-               free_commit_buffer(commit);
-
                if (!commit->parents && show_root)
                        printf("root %s\n", describe_object(&commit->object));
        }
@@ -352,29 +349,22 @@ static int fsck_obj(struct object *obj)
                struct tag *tag = (struct tag *) obj;
 
                if (show_tags && tag->tagged) {
-                       printf("tagged %s %s", typename(tag->tagged->type),
+                       printf("tagged %s %s", printable_type(tag->tagged),
                                describe_object(tag->tagged));
                        printf(" (%s) in %s\n", tag->tag,
                                describe_object(&tag->object));
                }
        }
 
-       return 0;
-}
-
-static int fsck_sha1(const unsigned char *sha1)
-{
-       struct object *obj = parse_object(sha1);
-       if (!obj) {
-               errors_found |= ERROR_OBJECT;
-               return error("%s: object corrupt or missing",
-                            sha1_to_hex(sha1));
-       }
-       obj->flags |= HAS_OBJ;
-       return fsck_obj(obj);
+out:
+       if (obj->type == OBJ_TREE)
+               free_tree_buffer((struct tree *)obj);
+       if (obj->type == OBJ_COMMIT)
+               free_commit_buffer((struct commit *)obj);
+       return err;
 }
 
-static int fsck_obj_buffer(const unsigned char *sha1, enum object_type type,
+static int fsck_obj_buffer(const struct object_id *oid, enum object_type type,
                           unsigned long size, void *buffer, int *eaten)
 {
        /*
@@ -382,50 +372,51 @@ static int fsck_obj_buffer(const unsigned char *sha1, enum object_type type,
         * verify_packfile(), data_valid variable for details.
         */
        struct object *obj;
-       obj = parse_object_buffer(sha1, type, size, buffer, eaten);
+       obj = parse_object_buffer(oid, type, size, buffer, eaten);
        if (!obj) {
                errors_found |= ERROR_OBJECT;
-               return error("%s: object corrupt or missing", sha1_to_hex(sha1));
+               return error("%s: object corrupt or missing", oid_to_hex(oid));
        }
-       obj->flags = HAS_OBJ;
+       obj->flags &= ~(REACHABLE | SEEN);
+       obj->flags |= HAS_OBJ;
        return fsck_obj(obj);
 }
 
 static int default_refs;
 
-static void fsck_handle_reflog_sha1(const char *refname, unsigned char *sha1,
-       unsigned long timestamp)
+static void fsck_handle_reflog_oid(const char *refname, struct object_id *oid,
+       timestamp_t timestamp)
 {
        struct object *obj;
 
-       if (!is_null_sha1(sha1)) {
-               obj = lookup_object(sha1);
-               if (obj) {
+       if (!is_null_oid(oid)) {
+               obj = lookup_object(oid->hash);
+               if (obj && (obj->flags & HAS_OBJ)) {
                        if (timestamp && name_objects)
                                add_decoration(fsck_walk_options.object_names,
                                        obj,
-                                       xstrfmt("%s@{%ld}", refname, timestamp));
-                       obj->used = 1;
+                                       xstrfmt("%s@{%"PRItime"}", refname, timestamp));
+                       obj->flags |= USED;
                        mark_object_reachable(obj);
                } else {
-                       error("%s: invalid reflog entry %s", refname, sha1_to_hex(sha1));
+                       error("%s: invalid reflog entry %s", refname, oid_to_hex(oid));
                        errors_found |= ERROR_REACHABLE;
                }
        }
 }
 
-static int fsck_handle_reflog_ent(unsigned char *osha1, unsigned char *nsha1,
-               const char *email, unsigned long timestamp, int tz,
+static int fsck_handle_reflog_ent(struct object_id *ooid, struct object_id *noid,
+               const char *email, timestamp_t timestamp, int tz,
                const char *message, void *cb_data)
 {
        const char *refname = cb_data;
 
        if (verbose)
                fprintf(stderr, "Checking reflog %s->%s\n",
-                       sha1_to_hex(osha1), sha1_to_hex(nsha1));
+                       oid_to_hex(ooid), oid_to_hex(noid));
 
-       fsck_handle_reflog_sha1(refname, osha1, 0);
-       fsck_handle_reflog_sha1(refname, nsha1, timestamp);
+       fsck_handle_reflog_oid(refname, ooid, 0);
+       fsck_handle_reflog_oid(refname, noid, timestamp);
        return 0;
 }
 
@@ -441,7 +432,7 @@ static int fsck_handle_ref(const char *refname, const struct object_id *oid,
 {
        struct object *obj;
 
-       obj = parse_object(oid->hash);
+       obj = parse_object(oid);
        if (!obj) {
                error("%s: invalid sha1 pointer %s", refname, oid_to_hex(oid));
                errors_found |= ERROR_REACHABLE;
@@ -453,7 +444,7 @@ static int fsck_handle_ref(const char *refname, const struct object_id *oid,
                errors_found |= ERROR_REFS;
        }
        default_refs++;
-       obj->used = 1;
+       obj->flags |= USED;
        if (name_objects)
                add_decoration(fsck_walk_options.object_names,
                        obj, xstrdup(refname));
@@ -488,9 +479,42 @@ static void get_default_heads(void)
        }
 }
 
-static int fsck_loose(const unsigned char *sha1, const char *path, void *data)
+static struct object *parse_loose_object(const struct object_id *oid,
+                                        const char *path)
+{
+       struct object *obj;
+       void *contents;
+       enum object_type type;
+       unsigned long size;
+       int eaten;
+
+       if (read_loose_object(path, oid->hash, &type, &size, &contents) < 0)
+               return NULL;
+
+       if (!contents && type != OBJ_BLOB)
+               die("BUG: read_loose_object streamed a non-blob");
+
+       obj = parse_object_buffer(oid, type, size, contents, &eaten);
+
+       if (!eaten)
+               free(contents);
+       return obj;
+}
+
+static int fsck_loose(const struct object_id *oid, const char *path, void *data)
 {
-       if (fsck_sha1(sha1))
+       struct object *obj = parse_loose_object(oid, path);
+
+       if (!obj) {
+               errors_found |= ERROR_OBJECT;
+               error("%s: object corrupt or missing: %s",
+                     oid_to_hex(oid), path);
+               return 0; /* keep checking other objects */
+       }
+
+       obj->flags &= ~(REACHABLE | SEEN);
+       obj->flags |= HAS_OBJ;
+       if (fsck_obj(obj))
                errors_found |= ERROR_OBJECT;
        return 0;
 }
@@ -502,7 +526,7 @@ static int fsck_cruft(const char *basename, const char *path, void *data)
        return 0;
 }
 
-static int fsck_subdir(int nr, const char *path, void *progress)
+static int fsck_subdir(unsigned int nr, const char *path, void *progress)
 {
        display_progress(progress, nr + 1);
        return 0;
@@ -564,14 +588,14 @@ static int fsck_cache_tree(struct cache_tree *it)
                fprintf(stderr, "Checking cache tree\n");
 
        if (0 <= it->entry_count) {
-               struct object *obj = parse_object(it->sha1);
+               struct object *obj = parse_object(&it->oid);
                if (!obj) {
                        error("%s: invalid sha1 pointer in cache-tree",
-                             sha1_to_hex(it->sha1));
+                             oid_to_hex(&it->oid));
                        errors_found |= ERROR_REFS;
                        return 1;
                }
-               obj->used = 1;
+               obj->flags |= USED;
                if (name_objects)
                        add_decoration(fsck_walk_options.object_names,
                                obj, xstrdup(":"));
@@ -584,6 +608,29 @@ static int fsck_cache_tree(struct cache_tree *it)
        return err;
 }
 
+static void mark_object_for_connectivity(const struct object_id *oid)
+{
+       struct object *obj = lookup_unknown_object(oid->hash);
+       obj->flags |= HAS_OBJ;
+}
+
+static int mark_loose_for_connectivity(const struct object_id *oid,
+                                      const char *path,
+                                      void *data)
+{
+       mark_object_for_connectivity(oid);
+       return 0;
+}
+
+static int mark_packed_for_connectivity(const struct object_id *oid,
+                                       struct packed_git *pack,
+                                       uint32_t pos,
+                                       void *data)
+{
+       mark_object_for_connectivity(oid);
+       return 0;
+}
+
 static char const * const fsck_usage[] = {
        N_("git fsck [<options>] [<object>...]"),
        NULL
@@ -609,7 +656,7 @@ static struct option fsck_opts[] = {
 
 int cmd_fsck(int argc, const char **argv, const char *prefix)
 {
-       int i, heads;
+       int i;
        struct alternate_object_database *alt;
 
        errors_found = 0;
@@ -640,60 +687,64 @@ int cmd_fsck(int argc, const char **argv, const char *prefix)
        git_config(fsck_config, NULL);
 
        fsck_head_link();
-       if (!connectivity_only) {
+       if (connectivity_only) {
+               for_each_loose_object(mark_loose_for_connectivity, NULL, 0);
+               for_each_packed_object(mark_packed_for_connectivity, NULL, 0);
+       } else {
                fsck_object_dir(get_object_directory());
 
                prepare_alt_odb();
                for (alt = alt_odb_list; alt; alt = alt->next)
                        fsck_object_dir(alt->path);
-       }
 
-       if (check_full) {
-               struct packed_git *p;
-               uint32_t total = 0, count = 0;
-               struct progress *progress = NULL;
+               if (check_full) {
+                       struct packed_git *p;
+                       uint32_t total = 0, count = 0;
+                       struct progress *progress = NULL;
+
+                       prepare_packed_git();
 
-               prepare_packed_git();
+                       if (show_progress) {
+                               for (p = packed_git; p; p = p->next) {
+                                       if (open_pack_index(p))
+                                               continue;
+                                       total += p->num_objects;
+                               }
 
-               if (show_progress) {
+                               progress = start_progress(_("Checking objects"), total);
+                       }
                        for (p = packed_git; p; p = p->next) {
-                               if (open_pack_index(p))
-                                       continue;
-                               total += p->num_objects;
+                               /* verify gives error messages itself */
+                               if (verify_pack(p, fsck_obj_buffer,
+                                               progress, count))
+                                       errors_found |= ERROR_PACK;
+                               count += p->num_objects;
                        }
-
-                       progress = start_progress(_("Checking objects"), total);
+                       stop_progress(&progress);
                }
-               for (p = packed_git; p; p = p->next) {
-                       /* verify gives error messages itself */
-                       if (verify_pack(p, fsck_obj_buffer,
-                                       progress, count))
-                               errors_found |= ERROR_PACK;
-                       count += p->num_objects;
-               }
-               stop_progress(&progress);
        }
 
-       heads = 0;
        for (i = 0; i < argc; i++) {
                const char *arg = argv[i];
-               unsigned char sha1[20];
-               if (!get_sha1(arg, sha1)) {
-                       struct object *obj = lookup_object(sha1);
+               struct object_id oid;
+               if (!get_oid(arg, &oid)) {
+                       struct object *obj = lookup_object(oid.hash);
 
-                       /* Error is printed by lookup_object(). */
-                       if (!obj)
+                       if (!obj || !(obj->flags & HAS_OBJ)) {
+                               error("%s: object missing", oid_to_hex(&oid));
+                               errors_found |= ERROR_OBJECT;
                                continue;
+                       }
 
-                       obj->used = 1;
+                       obj->flags |= USED;
                        if (name_objects)
                                add_decoration(fsck_walk_options.object_names,
                                        obj, xstrdup(arg));
                        mark_object_reachable(obj);
-                       heads++;
                        continue;
                }
                error("invalid parameter: expected sha1, got '%s'", arg);
+               errors_found |= ERROR_OBJECT;
        }
 
        /*
@@ -701,12 +752,13 @@ int cmd_fsck(int argc, const char **argv, const char *prefix)
         * default ones from .git/refs. We also consider the index file
         * in this case (ie this implies --cache).
         */
-       if (!heads) {
+       if (!argc) {
                get_default_heads();
                keep_cache_objects = 1;
        }
 
        if (keep_cache_objects) {
+               verify_index_checksum = 1;
                read_cache();
                for (i = 0; i < active_nr; i++) {
                        unsigned int mode;
@@ -716,11 +768,11 @@ int cmd_fsck(int argc, const char **argv, const char *prefix)
                        mode = active_cache[i]->ce_mode;
                        if (S_ISGITLINK(mode))
                                continue;
-                       blob = lookup_blob(active_cache[i]->oid.hash);
+                       blob = lookup_blob(&active_cache[i]->oid);
                        if (!blob)
                                continue;
                        obj = &blob->object;
-                       obj->used = 1;
+                       obj->flags |= USED;
                        if (name_objects)
                                add_decoration(fsck_walk_options.object_names,
                                        obj,