fsck-cache: sort entries by inode number
[gitweb.git] / fsck-cache.c
index 5dca6dbc04b0bf40cd654f05766f7daa6fac75b4..cb010957479976b9d8fdec6dc76e7a6c30e83a1b 100644 (file)
-#include "cache.h"
-
 #include <sys/types.h>
 #include <dirent.h>
 
-struct needs {
-       unsigned char parent[20];
-       unsigned char needs[20];
-       char tag[10];
-};
-
-struct seen {
-       unsigned char sha1[20];
-       char tag[10];
-       unsigned needed;
-};
-
-static struct needs *needs;
-static struct seen *seen;
-
-static int nr_seen, alloc_seen, nr_needs, alloc_needs;
-
-/*
- * These two functions build up a graph in memory about
- * what objects we've referenced, and found, and types..
- */
-static int compare_seen(const void *s1, const void *s2)
-{
-       return memcmp(s1, s2, 20);
-}
-
-static int lookup_seen(unsigned char *sha1, char *tag)
-{
-       int first = 0, last = nr_seen;
+#include "cache.h"
+#include "commit.h"
+#include "tree.h"
+#include "blob.h"
+#include "tag.h"
 
-       while (last > first) {
-               int next = (last + first) / 2;
-               struct seen *s = seen + next;
-               int cmp = memcmp(sha1, s->sha1, 20);
+#define REACHABLE 0x0001
 
-               if (cmp < 0) {
-                       last = next;
-                       continue;
-               }
-               if (cmp > 0) {
-                       first = next+1;
-                       continue;
-               }
-               if (strcmp(tag, s->tag))
-                       break;
-               s->needed++;
-               return 1;
-       }
-       return 0;
-}
+static int show_root = 0;
+static int show_tags = 0;
+static int show_unreachable = 0;
+static unsigned char head_sha1[20];
 
 static void check_connectivity(void)
 {
        int i;
 
-       /* Sort the "seen" tags for quicker lookup */
-       qsort(seen, nr_seen, sizeof(struct seen), compare_seen);
-
        /* Look up all the requirements, warn about missing objects.. */
-       for (i = 0; i < nr_needs; i++) {
-               struct needs *n = needs + i;
-               char hex[60];
+       for (i = 0; i < nr_objs; i++) {
+               struct object *obj = objs[i];
+               struct object_list *refs;
 
-               if (lookup_seen(n->needs, n->tag))
+               if (!obj->parsed) {
+                       printf("missing %s %s\n", obj->type, sha1_to_hex(obj->sha1));
                        continue;
-               strcpy(hex, sha1_to_hex(n->parent));
-       }
+               }
 
-       /* Tell the user about things not referenced.. */
-       for (i = 0; i < nr_seen; i++) {
-               struct seen *s = seen + i;
+               for (refs = obj->refs; refs; refs = refs->next) {
+                       if (refs->item->parsed)
+                               continue;
+                       printf("broken link from %7s %s\n",
+                              obj->type, sha1_to_hex(obj->sha1));
+                       printf("              to %7s %s\n",
+                              obj->type, sha1_to_hex(refs->item->sha1));
+               }
 
-               if (s->needed)
+               /* Don't bother with tag reachability. */
+               if (obj->type == tag_type)
                        continue;
-               printf("unreferenced %s: %s\n", s->tag, sha1_to_hex(s->sha1));
-       }
-}
 
-static void mark_needs_sha1(unsigned char *parent, const char * tag, unsigned char *child)
-{
-       struct needs *n;
+               if (show_unreachable && !(obj->flags & REACHABLE)) {
+                       printf("unreachable %s %s\n", obj->type, sha1_to_hex(obj->sha1));
+                       continue;
+               }
 
-       if (nr_needs == alloc_needs) {
-               alloc_needs = alloc_nr(alloc_needs);
-               needs = realloc(needs, alloc_needs*sizeof(struct needs));
+               if (!obj->used) {
+                       printf("dangling %s %s\n", obj->type, 
+                              sha1_to_hex(obj->sha1));
+               }
        }
-       n = needs + nr_needs;
-       nr_needs++;
-       memcpy(n->parent, parent, 20);
-       memcpy(n->needs, child, 20);
-       strncpy(n->tag, tag, sizeof(n->tag));
 }
 
-static int mark_sha1_seen(unsigned char *sha1, char *tag)
+static int fsck_tree(struct tree *item)
 {
-       struct seen *s;
-
-       if (nr_seen == alloc_seen) {
-               alloc_seen = alloc_nr(alloc_seen);
-               seen = realloc(seen, alloc_seen*sizeof(struct seen));
+       if (item->has_full_path) {
+               fprintf(stderr, "warning: fsck-cache: tree %s "
+                       "has full pathnames in it\n", 
+                       sha1_to_hex(item->object.sha1));
        }
-       s = seen + nr_seen;
-       memset(s, 0, sizeof(s));
-       nr_seen++;
-       memcpy(s->sha1, sha1, 20);
-       strncpy(s->tag, tag, sizeof(s->tag));
-       
        return 0;
 }
 
-static int fsck_tree(unsigned char *sha1, void *data, unsigned long size)
+static int fsck_commit(struct commit *commit)
 {
-       int warn_old_tree = 1;
-
-       while (size) {
-               int len = 1+strlen(data);
-               unsigned char *file_sha1 = data + len;
-               char *path = strchr(data, ' ');
-               unsigned int mode;
-               if (size < len + 20 || !path || sscanf(data, "%o", &mode) != 1)
-                       return -1;
-
-               /* Warn about trees that don't do the recursive thing.. */
-               if (warn_old_tree && strchr(path, '/')) {
-                       fprintf(stderr, "warning: fsck-cache: tree %s has full pathnames in it\n", sha1_to_hex(sha1));
-                       warn_old_tree = 0;
-               }
-
-               data += len + 20;
-               size -= len + 20;
-               mark_needs_sha1(sha1, S_ISDIR(mode) ? "tree" : "blob", file_sha1);
-       }
+       if (!commit->tree)
+               return -1;
+       if (!commit->parents && show_root)
+               printf("root %s\n", sha1_to_hex(commit->object.sha1));
+       if (!commit->date)
+               printf("bad commit date in %s\n", 
+                      sha1_to_hex(commit->object.sha1));
        return 0;
 }
 
-static int fsck_commit(unsigned char *sha1, void *data, unsigned long size)
+static int fsck_tag(struct tag *tag)
 {
-       int parents;
-       unsigned char tree_sha1[20];
-       unsigned char parent_sha1[20];
-
-       if (memcmp(data, "tree ", 5))
-               return -1;
-       if (get_sha1_hex(data + 5, tree_sha1) < 0)
-               return -1;
-       mark_needs_sha1(sha1, "tree", tree_sha1);
-       data += 5 + 40 + 1;     /* "tree " + <hex sha1> + '\n' */
-       parents = 0;
-       while (!memcmp(data, "parent ", 7)) {
-               if (get_sha1_hex(data + 7, parent_sha1) < 0)
-                       return -1;
-               mark_needs_sha1(sha1, "commit", parent_sha1);
-               data += 7 + 40 + 1;     /* "parent " + <hex sha1> + '\n' */
-               parents++;
-       }
-       if (!parents)
-               printf("root: %s\n", sha1_to_hex(sha1));
+       if (!show_tags)
+               return 0;
+
+       printf("tagged %s %s",
+              tag->tagged->type,
+              sha1_to_hex(tag->tagged->sha1));
+       printf(" (%s) in %s\n",
+              tag->tag, sha1_to_hex(tag->object.sha1));
        return 0;
 }
 
-static int fsck_entry(unsigned char *sha1, char *tag, void *data, unsigned long size)
+static int fsck_sha1(unsigned char *sha1)
 {
-       if (!strcmp(tag, "blob")) {
-               /* Nothing to check */;
-       } else if (!strcmp(tag, "tree")) {
-               if (fsck_tree(sha1, data, size) < 0)
-                       return -1;
-       } else if (!strcmp(tag, "commit")) {
-               if (fsck_commit(sha1, data, size) < 0)
-                       return -1;
-       } else
+       struct object *obj = parse_object(sha1);
+       if (!obj)
                return -1;
-       return mark_sha1_seen(sha1, tag);
+       if (obj->type == blob_type)
+               return 0;
+       if (obj->type == tree_type)
+               return fsck_tree((struct tree *) obj);
+       if (obj->type == commit_type)
+               return fsck_commit((struct commit *) obj);
+       if (obj->type == tag_type)
+               return fsck_tag((struct tag *) obj);
+       return -1;
 }
 
-static int fsck_name(char *hex)
-{
+/*
+ * This is the sorting chunk size: make it reasonably
+ * big so that we can sort well..
+ */
+#define MAX_SHA1_ENTRIES (1024)
+
+struct sha1_entry {
+       unsigned long ino;
        unsigned char sha1[20];
-       if (!get_sha1_hex(hex, sha1)) {
-               unsigned long mapsize;
-               void *map = map_sha1_file(sha1, &mapsize);
-               if (map) {
-                       char type[100];
-                       unsigned long size;
-                       void *buffer = NULL;
-                       if (!check_sha1_signature(sha1, map, mapsize))
-                               buffer = unpack_sha1_file(map, mapsize, type, &size);
-                       munmap(map, mapsize);
-                       if (buffer && !fsck_entry(sha1, type, buffer, size))
-                               return 0;
-               }
+};
+
+static struct {
+       unsigned long nr;
+       struct sha1_entry *entry[MAX_SHA1_ENTRIES];
+} sha1_list;
+
+static int ino_compare(const void *_a, const void *_b)
+{
+       const struct sha1_entry *a = _a, *b = _b;
+       unsigned long ino1 = a->ino, ino2 = b->ino;
+       return ino1 < ino2 ? -1 : ino1 > ino2 ? 1 : 0;
+}
+
+static void fsck_sha1_list(void)
+{
+       int i, nr = sha1_list.nr;
+
+       qsort(sha1_list.entry, nr, sizeof(struct sha1_entry *), ino_compare);
+       for (i = 0; i < nr; i++) {
+               struct sha1_entry *entry = sha1_list.entry[i];
+               unsigned char *sha1 = entry->sha1;
+
+               sha1_list.entry[i] = NULL;
+               if (fsck_sha1(sha1) < 0)
+                       fprintf(stderr, "bad sha1 entry '%s'\n", sha1_to_hex(sha1));
+               free(entry);
        }
-       return -1;
+       sha1_list.nr = 0;
+}
+
+static void add_sha1_list(unsigned char *sha1, unsigned long ino)
+{
+       struct sha1_entry *entry = xmalloc(sizeof(*entry));
+       int nr;
+
+       entry->ino = ino;
+       memcpy(entry->sha1, sha1, 20);
+       nr = sha1_list.nr;
+       if (nr == MAX_SHA1_ENTRIES) {
+               fsck_sha1_list();
+               nr = 0;
+       }
+       sha1_list.entry[nr] = entry;
+       sha1_list.nr = ++nr;
 }
 
 static int fsck_dir(int i, char *path)
@@ -204,12 +166,12 @@ static int fsck_dir(int i, char *path)
        struct dirent *de;
 
        if (!dir) {
-               fprintf(stderr, "missing sha1 directory '%s'", path);
-               return -1;
+               return error("missing sha1 directory '%s'", path);
        }
 
        while ((de = readdir(dir)) != NULL) {
                char name[100];
+               unsigned char sha1[20];
                int len = strlen(de->d_name);
 
                switch (len) {
@@ -223,8 +185,10 @@ static int fsck_dir(int i, char *path)
                case 38:
                        sprintf(name, "%02x", i);
                        memcpy(name+2, de->d_name, len+1);
-                       if (!fsck_name(name))
-                               continue;
+                       if (get_sha1_hex(name, sha1) < 0)
+                               break;
+                       add_sha1_list(sha1, de->d_ino);
+                       continue;
                }
                fprintf(stderr, "bad sha1 file: %s/%s\n", path, de->d_name);
        }
@@ -234,17 +198,68 @@ static int fsck_dir(int i, char *path)
 
 int main(int argc, char **argv)
 {
-       int i;
+       int i, heads;
        char *sha1_dir;
 
-       if (argc != 1)
-               usage("fsck-cache");
+       for (i = 1; i < argc; i++) {
+               const char *arg = argv[i];
+
+               if (!strcmp(arg, "--unreachable")) {
+                       show_unreachable = 1;
+                       continue;
+               }
+               if (!strcmp(arg, "--tags")) {
+                       show_tags = 1;
+                       continue;
+               }
+               if (!strcmp(arg, "--root")) {
+                       show_root = 1;
+                       continue;
+               }
+               if (*arg == '-')
+                       usage("fsck-cache [--tags] [[--unreachable] <head-sha1>*]");
+       }
+
        sha1_dir = getenv(DB_ENVIRONMENT) ? : DEFAULT_DB_ENVIRONMENT;
        for (i = 0; i < 256; i++) {
                static char dir[4096];
                sprintf(dir, "%s/%02x", sha1_dir, i);
                fsck_dir(i, dir);
        }
+       fsck_sha1_list();
+
+       heads = 0;
+       for (i = 1; i < argc; i++) {
+               const char *arg = argv[i]; 
+
+               if (*arg == '-')
+                       continue;
+
+               if (!get_sha1(arg, head_sha1)) {
+                       struct commit *commit = lookup_commit(head_sha1);
+                       struct object *obj;
+
+                       /* Error is printed by lookup_commit(). */
+                       if (!commit)
+                               continue;
+
+                       obj = &commit->object;
+                       obj->used = 1;
+                       mark_reachable(obj, REACHABLE);
+                       heads++;
+                       continue;
+               }
+               error("expected sha1, got %s", arg);
+       }
+
+       if (!heads) {
+               if (show_unreachable) {
+                       fprintf(stderr, "unable to do reachability without a head\n");
+                       show_unreachable = 0; 
+               }
+               fprintf(stderr, "expect dangling commits - potential heads - due to lack of head information\n");
+       }
+
        check_connectivity();
        return 0;
 }