Code

git-fsck-objects is now synonym to git-fsck
authorJunio C Hamano <junkio@cox.net>
Mon, 29 Jan 2007 00:33:58 +0000 (16:33 -0800)
committerJunio C Hamano <junkio@cox.net>
Mon, 29 Jan 2007 00:33:58 +0000 (16:33 -0800)
Signed-off-by: Junio C Hamano <junkio@cox.net>
16 files changed:
.gitignore
Documentation/cmd-list.perl
Documentation/core-intro.txt
Documentation/everyday.txt
Documentation/git-fsck-objects.txt
Documentation/git-fsck.txt [new file with mode: 0644]
Documentation/git-pack-redundant.txt
Documentation/git-prune.txt
Makefile
fsck-objects.c [deleted file]
fsck.c [new file with mode: 0644]
git-lost-found.sh
http-fetch.c
t/t1410-reflog.sh
t/t5500-fetch-pack.sh
t/t5710-info-alternate.sh

index b4dccd7b816e958273ec84b77e8ff4d1602461de..a43444fb92e3475eaadbc7792e0defb9153910fb 100644 (file)
@@ -42,6 +42,7 @@ git-findtags
 git-fmt-merge-msg
 git-for-each-ref
 git-format-patch
+git-fsck
 git-fsck-objects
 git-gc
 git-get-tar-commit-id
index 824462551f4663fcce2d97b0dccbcba0851a5bdf..6dba8d8fe05769efa69ccc97ebdcbb2af0ca56e2 100755 (executable)
@@ -97,7 +97,7 @@ git-fetch-pack                          synchingrepositories
 git-fmt-merge-msg                       purehelpers
 git-for-each-ref                        plumbinginterrogators
 git-format-patch                        mainporcelain
-git-fsck-objects                        ancillaryinterrogators
+git-fsck                               ancillaryinterrogators
 git-gc                                  mainporcelain
 git-get-tar-commit-id                   ancillaryinterrogators
 git-grep                                mainporcelain
index 0458dc3d0263d40854e07d1a8ac53d1ec88379e7..abafefc71c37d8a3015f0227d24da0f07096fe4c 100644 (file)
@@ -82,7 +82,7 @@ size> + <byte\0> + <binary object data>.
 
 The structured objects can further have their structure and
 connectivity to other objects verified. This is generally done with
-the `git-fsck-objects` program, which generates a full dependency graph
+the `git-fsck` program, which generates a full dependency graph
 of all objects, and verifies their internal consistency (in addition
 to just verifying their superficial consistency through the hash).
 
index fbbbc92b811f32581e20f96c3c664368ce4e801b..08c61b1f1ac6ca880c6c3a4f311f65486aabd23e 100644 (file)
@@ -28,7 +28,7 @@ Everybody uses these commands to maintain git repositories.
   * gitlink:git-init[1] or gitlink:git-clone[1] to create a
     new repository.
 
-  * gitlink:git-fsck-objects[1] to check the repository for errors.
+  * gitlink:git-fsck[1] to check the repository for errors.
 
   * gitlink:git-prune[1] to remove unused objects in the repository.
 
@@ -43,7 +43,7 @@ Examples
 Check health and remove cruft.::
 +
 ------------
-$ git fsck-objects <1>
+$ git fsck <1>
 $ git count-objects <2>
 $ git repack <3>
 $ git gc <4>
index d0af99d3512d3794f9907612cef1d30616120064..f21061ecfe56a438238cc84761174f836bd0035e 100644 (file)
@@ -8,132 +8,10 @@ git-fsck-objects - Verifies the connectivity and validity of the objects in the
 
 SYNOPSIS
 --------
-[verse]
-'git-fsck-objects' [--tags] [--root] [--unreachable] [--cache]
-                [--full] [--strict] [<object>*]
+'git-fsck-objects' ...
 
 DESCRIPTION
 -----------
-Verifies the connectivity and validity of the objects in the database.
-
-OPTIONS
--------
-<object>::
-       An object to treat as the head of an unreachability trace.
-+
-If no objects are given, git-fsck-objects defaults to using the
-index file and all SHA1 references in .git/refs/* as heads.
-
---unreachable::
-       Print out objects that exist but that aren't readable from any
-       of the reference nodes.
-
---root::
-       Report root nodes.
-
---tags::
-       Report tags.
-
---cache::
-       Consider any object recorded in the index also as a head node for
-       an unreachability trace.
-
---full::
-       Check not just objects in GIT_OBJECT_DIRECTORY
-       ($GIT_DIR/objects), but also the ones found in alternate
-       object pools listed in GIT_ALTERNATE_OBJECT_DIRECTORIES
-       or $GIT_DIR/objects/info/alternates,
-       and in packed git archives found in $GIT_DIR/objects/pack
-       and corresponding pack subdirectories in alternate
-       object pools.
-
---strict::
-       Enable more strict checking, namely to catch a file mode
-       recorded with g+w bit set, which was created by older
-       versions of git.  Existing repositories, including the
-       Linux kernel, git itself, and sparse repository have old
-       objects that triggers this check, but it is recommended
-       to check new projects with this flag.
-
-It tests SHA1 and general object sanity, and it does full tracking of
-the resulting reachability and everything else. It prints out any
-corruption it finds (missing or bad objects), and if you use the
-'--unreachable' flag it will also print out objects that exist but
-that aren't readable from any of the specified head nodes.
-
-So for example
-
-       git-fsck-objects --unreachable HEAD $(cat .git/refs/heads/*)
-
-will do quite a _lot_ of verification on the tree. There are a few
-extra validity tests to be added (make sure that tree objects are
-sorted properly etc), but on the whole if "git-fsck-objects" is happy, you
-do have a valid tree.
-
-Any corrupt objects you will have to find in backups or other archives
-(i.e., you can just remove them and do an "rsync" with some other site in
-the hopes that somebody else has the object you have corrupted).
-
-Of course, "valid tree" doesn't mean that it wasn't generated by some
-evil person, and the end result might be crap. git is a revision
-tracking system, not a quality assurance system ;)
-
-Extracted Diagnostics
----------------------
-
-expect dangling commits - potential heads - due to lack of head information::
-       You haven't specified any nodes as heads so it won't be
-       possible to differentiate between un-parented commits and
-       root nodes.
-
-missing sha1 directory '<dir>'::
-       The directory holding the sha1 objects is missing.
-
-unreachable <type> <object>::
-       The <type> object <object>, isn't actually referred to directly
-       or indirectly in any of the trees or commits seen. This can
-       mean that there's another root node that you're not specifying
-       or that the tree is corrupt. If you haven't missed a root node
-       then you might as well delete unreachable nodes since they
-       can't be used.
-
-missing <type> <object>::
-       The <type> object <object>, is referred to but isn't present in
-       the database.
-
-dangling <type> <object>::
-       The <type> object <object>, is present in the database but never
-       'directly' used. A dangling commit could be a root node.
-
-warning: git-fsck-objects: tree <tree> has full pathnames in it::
-       And it shouldn't...
-
-sha1 mismatch <object>::
-       The database has an object who's sha1 doesn't match the
-       database value.
-       This indicates a serious data integrity problem.
-
-Environment Variables
----------------------
-
-GIT_OBJECT_DIRECTORY::
-       used to specify the object database root (usually $GIT_DIR/objects)
-
-GIT_INDEX_FILE::
-       used to specify the index file of the index
-
-GIT_ALTERNATE_OBJECT_DIRECTORIES::
-       used to specify additional object database roots (usually unset)
-
-Author
-------
-Written by Linus Torvalds <torvalds@osdl.org>
-
-Documentation
---------------
-Documentation by David Greaves, Junio C Hamano and the git-list <git@vger.kernel.org>.
-
-GIT
----
-Part of the gitlink:git[7] suite
 
+This is a synonym for gitlink:git-fsck[1].  Please refer to the
+documentation of that command.
diff --git a/Documentation/git-fsck.txt b/Documentation/git-fsck.txt
new file mode 100644 (file)
index 0000000..058009d
--- /dev/null
@@ -0,0 +1,139 @@
+git-fsck(1)
+===========
+
+NAME
+----
+git-fsck - Verifies the connectivity and validity of the objects in the database
+
+
+SYNOPSIS
+--------
+[verse]
+'git-fsck' [--tags] [--root] [--unreachable] [--cache]
+                [--full] [--strict] [<object>*]
+
+DESCRIPTION
+-----------
+Verifies the connectivity and validity of the objects in the database.
+
+OPTIONS
+-------
+<object>::
+       An object to treat as the head of an unreachability trace.
++
+If no objects are given, git-fsck defaults to using the
+index file and all SHA1 references in .git/refs/* as heads.
+
+--unreachable::
+       Print out objects that exist but that aren't readable from any
+       of the reference nodes.
+
+--root::
+       Report root nodes.
+
+--tags::
+       Report tags.
+
+--cache::
+       Consider any object recorded in the index also as a head node for
+       an unreachability trace.
+
+--full::
+       Check not just objects in GIT_OBJECT_DIRECTORY
+       ($GIT_DIR/objects), but also the ones found in alternate
+       object pools listed in GIT_ALTERNATE_OBJECT_DIRECTORIES
+       or $GIT_DIR/objects/info/alternates,
+       and in packed git archives found in $GIT_DIR/objects/pack
+       and corresponding pack subdirectories in alternate
+       object pools.
+
+--strict::
+       Enable more strict checking, namely to catch a file mode
+       recorded with g+w bit set, which was created by older
+       versions of git.  Existing repositories, including the
+       Linux kernel, git itself, and sparse repository have old
+       objects that triggers this check, but it is recommended
+       to check new projects with this flag.
+
+It tests SHA1 and general object sanity, and it does full tracking of
+the resulting reachability and everything else. It prints out any
+corruption it finds (missing or bad objects), and if you use the
+'--unreachable' flag it will also print out objects that exist but
+that aren't readable from any of the specified head nodes.
+
+So for example
+
+       git-fsck --unreachable HEAD $(cat .git/refs/heads/*)
+
+will do quite a _lot_ of verification on the tree. There are a few
+extra validity tests to be added (make sure that tree objects are
+sorted properly etc), but on the whole if "git-fsck" is happy, you
+do have a valid tree.
+
+Any corrupt objects you will have to find in backups or other archives
+(i.e., you can just remove them and do an "rsync" with some other site in
+the hopes that somebody else has the object you have corrupted).
+
+Of course, "valid tree" doesn't mean that it wasn't generated by some
+evil person, and the end result might be crap. git is a revision
+tracking system, not a quality assurance system ;)
+
+Extracted Diagnostics
+---------------------
+
+expect dangling commits - potential heads - due to lack of head information::
+       You haven't specified any nodes as heads so it won't be
+       possible to differentiate between un-parented commits and
+       root nodes.
+
+missing sha1 directory '<dir>'::
+       The directory holding the sha1 objects is missing.
+
+unreachable <type> <object>::
+       The <type> object <object>, isn't actually referred to directly
+       or indirectly in any of the trees or commits seen. This can
+       mean that there's another root node that you're not specifying
+       or that the tree is corrupt. If you haven't missed a root node
+       then you might as well delete unreachable nodes since they
+       can't be used.
+
+missing <type> <object>::
+       The <type> object <object>, is referred to but isn't present in
+       the database.
+
+dangling <type> <object>::
+       The <type> object <object>, is present in the database but never
+       'directly' used. A dangling commit could be a root node.
+
+warning: git-fsck: tree <tree> has full pathnames in it::
+       And it shouldn't...
+
+sha1 mismatch <object>::
+       The database has an object who's sha1 doesn't match the
+       database value.
+       This indicates a serious data integrity problem.
+
+Environment Variables
+---------------------
+
+GIT_OBJECT_DIRECTORY::
+       used to specify the object database root (usually $GIT_DIR/objects)
+
+GIT_INDEX_FILE::
+       used to specify the index file of the index
+
+GIT_ALTERNATE_OBJECT_DIRECTORIES::
+       used to specify additional object database roots (usually unset)
+
+Author
+------
+Written by Linus Torvalds <torvalds@osdl.org>
+
+Documentation
+--------------
+Documentation by David Greaves, Junio C Hamano and the git-list <git@vger.kernel.org>.
+
+GIT
+---
+Part of the gitlink:git[7] suite
+
index be2c64eb57cdbc10c12c9784134911f91d791f8c..94bbea0db2caf32635cfa8fc199e8150b1cd05f2 100644 (file)
@@ -21,7 +21,7 @@ given will be ignored when checking which packs are required. This makes the
 following command useful when wanting to remove packs which contain unreachable
 objects.
 
-git-fsck-objects --full --unreachable | cut -d ' ' -f3 | \
+git-fsck --full --unreachable | cut -d ' ' -f3 | \
 git-pack-redundant --all | xargs rm
 
 OPTIONS
index a11e3030943ab3ffd10afa017640826795809d05..0b44f3015d1ce8c600eb274a0f74c53868949658 100644 (file)
@@ -13,7 +13,7 @@ SYNOPSIS
 DESCRIPTION
 -----------
 
-This runs `git-fsck-objects --unreachable` using all the refs
+This runs `git-fsck --unreachable` using all the refs
 available in `$GIT_DIR/refs`, optionally with additional set of
 objects specified on the command line, and prunes all
 objects unreachable from any of these head objects from the object database.
index fd8a522efa110d9d2b178201dd9b66bffda7a46f..1552b2a5308f0a7c73b22f3416d5268e75a23deb 100644 (file)
--- a/Makefile
+++ b/Makefile
@@ -192,7 +192,7 @@ SCRIPTS = $(patsubst %.sh,%,$(SCRIPT_SH)) \
 
 # ... and all the rest that could be moved out of bindir to gitexecdir
 PROGRAMS = \
-       git-convert-objects$X git-fetch-pack$X git-fsck-objects$X \
+       git-convert-objects$X git-fetch-pack$X git-fsck$X \
        git-hash-object$X git-index-pack$X git-local-fetch$X \
        git-merge-base$X \
        git-daemon$X \
@@ -214,6 +214,7 @@ EXTRA_PROGRAMS =
 BUILT_INS = \
        git-format-patch$X git-show$X git-whatchanged$X git-cherry$X \
        git-get-tar-commit-id$X git-init$X git-repo-config$X \
+       git-fsck-objects$X \
        $(patsubst builtin-%.o,git-%$X,$(BUILTIN_OBJS))
 
 # what 'all' will build and 'install' will install, in gitexecdir
diff --git a/fsck-objects.c b/fsck-objects.c
deleted file mode 100644 (file)
index ecfb014..0000000
+++ /dev/null
@@ -1,690 +0,0 @@
-#include "cache.h"
-#include "commit.h"
-#include "tree.h"
-#include "blob.h"
-#include "tag.h"
-#include "refs.h"
-#include "pack.h"
-#include "cache-tree.h"
-#include "tree-walk.h"
-
-#define REACHABLE 0x0001
-#define SEEN      0x0002
-
-static int show_root;
-static int show_tags;
-static int show_unreachable;
-static int check_full;
-static int check_strict;
-static int keep_cache_objects;
-static unsigned char head_sha1[20];
-
-#ifdef NO_D_INO_IN_DIRENT
-#define SORT_DIRENT 0
-#define DIRENT_SORT_HINT(de) 0
-#else
-#define SORT_DIRENT 1
-#define DIRENT_SORT_HINT(de) ((de)->d_ino)
-#endif
-
-static void objreport(struct object *obj, const char *severity,
-                      const char *err, va_list params)
-{
-       fprintf(stderr, "%s in %s %s: ",
-               severity, typename(obj->type), sha1_to_hex(obj->sha1));
-       vfprintf(stderr, err, params);
-       fputs("\n", stderr);
-}
-
-static int objerror(struct object *obj, const char *err, ...)
-{
-       va_list params;
-       va_start(params, err);
-       objreport(obj, "error", err, params);
-       va_end(params);
-       return -1;
-}
-
-static int objwarning(struct object *obj, const char *err, ...)
-{
-       va_list params;
-       va_start(params, err);
-       objreport(obj, "warning", err, params);
-       va_end(params);
-       return -1;
-}
-
-/*
- * Check a single reachable object
- */
-static void check_reachable_object(struct object *obj)
-{
-       const struct object_refs *refs;
-
-       /*
-        * We obviously want the object to be parsed,
-        * except if it was in a pack-file and we didn't
-        * do a full fsck
-        */
-       if (!obj->parsed) {
-               if (has_sha1_file(obj->sha1))
-                       return; /* it is in pack - forget about it */
-               printf("missing %s %s\n", typename(obj->type), sha1_to_hex(obj->sha1));
-               return;
-       }
-
-       /*
-        * Check that everything that we try to reference is also good.
-        */
-       refs = lookup_object_refs(obj);
-       if (refs) {
-               unsigned j;
-               for (j = 0; j < refs->count; j++) {
-                       struct object *ref = refs->ref[j];
-                       if (ref->parsed ||
-                           (has_sha1_file(ref->sha1)))
-                               continue;
-                       printf("broken link from %7s %s\n",
-                              typename(obj->type), sha1_to_hex(obj->sha1));
-                       printf("              to %7s %s\n",
-                              typename(ref->type), sha1_to_hex(ref->sha1));
-               }
-       }
-}
-
-/*
- * Check a single unreachable object
- */
-static void check_unreachable_object(struct object *obj)
-{
-       /*
-        * Missing unreachable object? Ignore it. It's not like
-        * we miss it (since it can't be reached), nor do we want
-        * to complain about it being unreachable (since it does
-        * not exist).
-        */
-       if (!obj->parsed)
-               return;
-
-       /*
-        * Unreachable object that exists? Show it if asked to,
-        * since this is something that is prunable.
-        */
-       if (show_unreachable) {
-               printf("unreachable %s %s\n", typename(obj->type), sha1_to_hex(obj->sha1));
-               return;
-       }
-
-       /*
-        * "!used" means that nothing at all points to it, including
-        * other unreacahble objects. In other words, it's the "tip"
-        * of some set of unreachable objects, usually a commit that
-        * got dropped.
-        *
-        * Such starting points are more interesting than some random
-        * set of unreachable objects, so we show them even if the user
-        * hasn't asked for _all_ unreachable objects. If you have
-        * deleted a branch by mistake, this is a prime candidate to
-        * start looking at, for example.
-        */
-       if (!obj->used) {
-               printf("dangling %s %s\n", typename(obj->type),
-                      sha1_to_hex(obj->sha1));
-               return;
-       }
-
-       /*
-        * Otherwise? It's there, it's unreachable, and some other unreachable
-        * object points to it. Ignore it - it's not interesting, and we showed
-        * all the interesting cases above.
-        */
-}
-
-static void check_object(struct object *obj)
-{
-       if (obj->flags & REACHABLE)
-               check_reachable_object(obj);
-       else
-               check_unreachable_object(obj);
-}
-
-static void check_connectivity(void)
-{
-       int i, max;
-
-       /* Look up all the requirements, warn about missing objects.. */
-       max = get_max_object_index();
-       for (i = 0; i < max; i++) {
-               struct object *obj = get_indexed_object(i);
-
-               if (obj)
-                       check_object(obj);
-       }
-}
-
-/*
- * The entries in a tree are ordered in the _path_ order,
- * which means that a directory entry is ordered by adding
- * a slash to the end of it.
- *
- * So a directory called "a" is ordered _after_ a file
- * called "a.c", because "a/" sorts after "a.c".
- */
-#define TREE_UNORDERED (-1)
-#define TREE_HAS_DUPS  (-2)
-
-static int verify_ordered(unsigned mode1, const char *name1, unsigned mode2, const char *name2)
-{
-       int len1 = strlen(name1);
-       int len2 = strlen(name2);
-       int len = len1 < len2 ? len1 : len2;
-       unsigned char c1, c2;
-       int cmp;
-
-       cmp = memcmp(name1, name2, len);
-       if (cmp < 0)
-               return 0;
-       if (cmp > 0)
-               return TREE_UNORDERED;
-
-       /*
-        * Ok, the first <len> characters are the same.
-        * Now we need to order the next one, but turn
-        * a '\0' into a '/' for a directory entry.
-        */
-       c1 = name1[len];
-       c2 = name2[len];
-       if (!c1 && !c2)
-               /*
-                * git-write-tree used to write out a nonsense tree that has
-                * entries with the same name, one blob and one tree.  Make
-                * sure we do not have duplicate entries.
-                */
-               return TREE_HAS_DUPS;
-       if (!c1 && S_ISDIR(mode1))
-               c1 = '/';
-       if (!c2 && S_ISDIR(mode2))
-               c2 = '/';
-       return c1 < c2 ? 0 : TREE_UNORDERED;
-}
-
-static int fsck_tree(struct tree *item)
-{
-       int retval;
-       int has_full_path = 0;
-       int has_zero_pad = 0;
-       int has_bad_modes = 0;
-       int has_dup_entries = 0;
-       int not_properly_sorted = 0;
-       struct tree_desc desc;
-       unsigned o_mode;
-       const char *o_name;
-       const unsigned char *o_sha1;
-
-       desc.buf = item->buffer;
-       desc.size = item->size;
-
-       o_mode = 0;
-       o_name = NULL;
-       o_sha1 = NULL;
-       while (desc.size) {
-               unsigned mode;
-               const char *name;
-               const unsigned char *sha1;
-
-               sha1 = tree_entry_extract(&desc, &name, &mode);
-
-               if (strchr(name, '/'))
-                       has_full_path = 1;
-               has_zero_pad |= *(char *)desc.buf == '0';
-               update_tree_entry(&desc);
-
-               switch (mode) {
-               /*
-                * Standard modes..
-                */
-               case S_IFREG | 0755:
-               case S_IFREG | 0644:
-               case S_IFLNK:
-               case S_IFDIR:
-                       break;
-               /*
-                * This is nonstandard, but we had a few of these
-                * early on when we honored the full set of mode
-                * bits..
-                */
-               case S_IFREG | 0664:
-                       if (!check_strict)
-                               break;
-               default:
-                       has_bad_modes = 1;
-               }
-
-               if (o_name) {
-                       switch (verify_ordered(o_mode, o_name, mode, name)) {
-                       case TREE_UNORDERED:
-                               not_properly_sorted = 1;
-                               break;
-                       case TREE_HAS_DUPS:
-                               has_dup_entries = 1;
-                               break;
-                       default:
-                               break;
-                       }
-               }
-
-               o_mode = mode;
-               o_name = name;
-               o_sha1 = sha1;
-       }
-       free(item->buffer);
-       item->buffer = NULL;
-
-       retval = 0;
-       if (has_full_path) {
-               objwarning(&item->object, "contains full pathnames");
-       }
-       if (has_zero_pad) {
-               objwarning(&item->object, "contains zero-padded file modes");
-       }
-       if (has_bad_modes) {
-               objwarning(&item->object, "contains bad file modes");
-       }
-       if (has_dup_entries) {
-               retval = objerror(&item->object, "contains duplicate file entries");
-       }
-       if (not_properly_sorted) {
-               retval = objerror(&item->object, "not properly sorted");
-       }
-       return retval;
-}
-
-static int fsck_commit(struct commit *commit)
-{
-       char *buffer = commit->buffer;
-       unsigned char tree_sha1[20], sha1[20];
-
-       if (memcmp(buffer, "tree ", 5))
-               return objerror(&commit->object, "invalid format - expected 'tree' line");
-       if (get_sha1_hex(buffer+5, tree_sha1) || buffer[45] != '\n')
-               return objerror(&commit->object, "invalid 'tree' line format - bad sha1");
-       buffer += 46;
-       while (!memcmp(buffer, "parent ", 7)) {
-               if (get_sha1_hex(buffer+7, sha1) || buffer[47] != '\n')
-                       return objerror(&commit->object, "invalid 'parent' line format - bad sha1");
-               buffer += 48;
-       }
-       if (memcmp(buffer, "author ", 7))
-               return objerror(&commit->object, "invalid format - expected 'author' line");
-       free(commit->buffer);
-       commit->buffer = NULL;
-       if (!commit->tree)
-               return objerror(&commit->object, "could not load commit's tree %s", tree_sha1);
-       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_tag(struct tag *tag)
-{
-       struct object *tagged = tag->tagged;
-
-       if (!tagged) {
-               return objerror(&tag->object, "could not load tagged object");
-       }
-       if (!show_tags)
-               return 0;
-
-       printf("tagged %s %s", typename(tagged->type), sha1_to_hex(tagged->sha1));
-       printf(" (%s) in %s\n", tag->tag, sha1_to_hex(tag->object.sha1));
-       return 0;
-}
-
-static int fsck_sha1(unsigned char *sha1)
-{
-       struct object *obj = parse_object(sha1);
-       if (!obj)
-               return error("%s: object corrupt or missing", sha1_to_hex(sha1));
-       if (obj->flags & SEEN)
-               return 0;
-       obj->flags |= SEEN;
-       if (obj->type == OBJ_BLOB)
-               return 0;
-       if (obj->type == OBJ_TREE)
-               return fsck_tree((struct tree *) obj);
-       if (obj->type == OBJ_COMMIT)
-               return fsck_commit((struct commit *) obj);
-       if (obj->type == OBJ_TAG)
-               return fsck_tag((struct tag *) obj);
-       /* By now, parse_object() would've returned NULL instead. */
-       return objerror(obj, "unknown type '%d' (internal fsck error)", obj->type);
-}
-
-/*
- * 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];
-};
-
-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;
-
-       if (SORT_DIRENT)
-               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;
-               fsck_sha1(sha1);
-               free(entry);
-       }
-       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;
-       hashcpy(entry->sha1, sha1);
-       nr = sha1_list.nr;
-       if (nr == MAX_SHA1_ENTRIES) {
-               fsck_sha1_list();
-               nr = 0;
-       }
-       sha1_list.entry[nr] = entry;
-       sha1_list.nr = ++nr;
-}
-
-static void fsck_dir(int i, char *path)
-{
-       DIR *dir = opendir(path);
-       struct dirent *de;
-
-       if (!dir)
-               return;
-
-       while ((de = readdir(dir)) != NULL) {
-               char name[100];
-               unsigned char sha1[20];
-               int len = strlen(de->d_name);
-
-               switch (len) {
-               case 2:
-                       if (de->d_name[1] != '.')
-                               break;
-               case 1:
-                       if (de->d_name[0] != '.')
-                               break;
-                       continue;
-               case 38:
-                       sprintf(name, "%02x", i);
-                       memcpy(name+2, de->d_name, len+1);
-                       if (get_sha1_hex(name, sha1) < 0)
-                               break;
-                       add_sha1_list(sha1, DIRENT_SORT_HINT(de));
-                       continue;
-               }
-               fprintf(stderr, "bad sha1 file: %s/%s\n", path, de->d_name);
-       }
-       closedir(dir);
-}
-
-static int default_refs;
-
-static int fsck_handle_reflog_ent(unsigned char *osha1, unsigned char *nsha1,
-               const char *email, unsigned long timestamp, int tz,
-               const char *message, void *cb_data)
-{
-       struct object *obj;
-
-       if (!is_null_sha1(osha1)) {
-               obj = lookup_object(osha1);
-               if (obj) {
-                       obj->used = 1;
-                       mark_reachable(obj, REACHABLE);
-               }
-       }
-       obj = lookup_object(nsha1);
-       if (obj) {
-               obj->used = 1;
-               mark_reachable(obj, REACHABLE);
-       }
-       return 0;
-}
-
-static int fsck_handle_ref(const char *refname, const unsigned char *sha1, int flag, void *cb_data)
-{
-       struct object *obj;
-
-       obj = lookup_object(sha1);
-       if (!obj) {
-               if (has_sha1_file(sha1)) {
-                       default_refs++;
-                       return 0; /* it is in a pack */
-               }
-               error("%s: invalid sha1 pointer %s", refname, sha1_to_hex(sha1));
-               /* We'll continue with the rest despite the error.. */
-               return 0;
-       }
-       default_refs++;
-       obj->used = 1;
-       mark_reachable(obj, REACHABLE);
-
-       for_each_reflog_ent(refname, fsck_handle_reflog_ent, NULL);
-
-       return 0;
-}
-
-static void get_default_heads(void)
-{
-       for_each_ref(fsck_handle_ref, NULL);
-
-       /*
-        * Not having any default heads isn't really fatal, but
-        * it does mean that "--unreachable" no longer makes any
-        * sense (since in this case everything will obviously
-        * be unreachable by definition.
-        *
-        * Showing dangling objects is valid, though (as those
-        * dangling objects are likely lost heads).
-        *
-        * So we just print a warning about it, and clear the
-        * "show_unreachable" flag.
-        */
-       if (!default_refs) {
-               error("No default references");
-               show_unreachable = 0;
-       }
-}
-
-static void fsck_object_dir(const char *path)
-{
-       int i;
-       for (i = 0; i < 256; i++) {
-               static char dir[4096];
-               sprintf(dir, "%s/%02x", path, i);
-               fsck_dir(i, dir);
-       }
-       fsck_sha1_list();
-}
-
-static int fsck_head_link(void)
-{
-       unsigned char sha1[20];
-       int flag;
-       const char *head_points_at = resolve_ref("HEAD", sha1, 1, &flag);
-
-       if (!head_points_at || !(flag & REF_ISSYMREF))
-               return error("HEAD is not a symbolic ref");
-       if (strncmp(head_points_at, "refs/heads/", 11))
-               return error("HEAD points to something strange (%s)",
-                            head_points_at);
-       if (is_null_sha1(sha1))
-               return error("HEAD: not a valid git pointer");
-       return 0;
-}
-
-static int fsck_cache_tree(struct cache_tree *it)
-{
-       int i;
-       int err = 0;
-
-       if (0 <= it->entry_count) {
-               struct object *obj = parse_object(it->sha1);
-               if (!obj) {
-                       error("%s: invalid sha1 pointer in cache-tree",
-                             sha1_to_hex(it->sha1));
-                       return 1;
-               }
-               mark_reachable(obj, REACHABLE);
-               obj->used = 1;
-               if (obj->type != OBJ_TREE)
-                       err |= objerror(obj, "non-tree in cache-tree");
-       }
-       for (i = 0; i < it->subtree_nr; i++)
-               err |= fsck_cache_tree(it->down[i]->cache_tree);
-       return err;
-}
-
-int main(int argc, char **argv)
-{
-       int i, heads;
-
-       track_object_refs = 1;
-       setup_git_directory();
-
-       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 (!strcmp(arg, "--cache")) {
-                       keep_cache_objects = 1;
-                       continue;
-               }
-               if (!strcmp(arg, "--full")) {
-                       check_full = 1;
-                       continue;
-               }
-               if (!strcmp(arg, "--strict")) {
-                       check_strict = 1;
-                       continue;
-               }
-               if (*arg == '-')
-                       usage("git-fsck-objects [--tags] [--root] [[--unreachable] [--cache] [--full] [--strict] <head-sha1>*]");
-       }
-
-       fsck_head_link();
-       fsck_object_dir(get_object_directory());
-       if (check_full) {
-               struct alternate_object_database *alt;
-               struct packed_git *p;
-               prepare_alt_odb();
-               for (alt = alt_odb_list; alt; alt = alt->next) {
-                       char namebuf[PATH_MAX];
-                       int namelen = alt->name - alt->base;
-                       memcpy(namebuf, alt->base, namelen);
-                       namebuf[namelen - 1] = 0;
-                       fsck_object_dir(namebuf);
-               }
-               prepare_packed_git();
-               for (p = packed_git; p; p = p->next)
-                       /* verify gives error messages itself */
-                       verify_pack(p, 0);
-
-               for (p = packed_git; p; p = p->next) {
-                       int num = num_packed_objects(p);
-                       for (i = 0; i < num; i++) {
-                               unsigned char sha1[20];
-                               nth_packed_object_sha1(p, i, sha1);
-                               fsck_sha1(sha1);
-                       }
-               }
-       }
-
-       heads = 0;
-       for (i = 1; i < argc; i++) {
-               const char *arg = argv[i]; 
-
-               if (*arg == '-')
-                       continue;
-
-               if (!get_sha1(arg, head_sha1)) {
-                       struct object *obj = lookup_object(head_sha1);
-
-                       /* Error is printed by lookup_object(). */
-                       if (!obj)
-                               continue;
-
-                       obj->used = 1;
-                       mark_reachable(obj, REACHABLE);
-                       heads++;
-                       continue;
-               }
-               error("invalid parameter: expected sha1, got '%s'", arg);
-       }
-
-       /*
-        * If we've not been given any explicit head information, do the
-        * default ones from .git/refs. We also consider the index file
-        * in this case (ie this implies --cache).
-        */
-       if (!heads) {
-               get_default_heads();
-               keep_cache_objects = 1;
-       }
-
-       if (keep_cache_objects) {
-               int i;
-               read_cache();
-               for (i = 0; i < active_nr; i++) {
-                       struct blob *blob = lookup_blob(active_cache[i]->sha1);
-                       struct object *obj;
-                       if (!blob)
-                               continue;
-                       obj = &blob->object;
-                       obj->used = 1;
-                       mark_reachable(obj, REACHABLE);
-               }
-               if (active_cache_tree)
-                       fsck_cache_tree(active_cache_tree);
-       }
-
-       check_connectivity();
-       return 0;
-}
diff --git a/fsck.c b/fsck.c
new file mode 100644 (file)
index 0000000..558f0a6
--- /dev/null
+++ b/fsck.c
@@ -0,0 +1,690 @@
+#include "cache.h"
+#include "commit.h"
+#include "tree.h"
+#include "blob.h"
+#include "tag.h"
+#include "refs.h"
+#include "pack.h"
+#include "cache-tree.h"
+#include "tree-walk.h"
+
+#define REACHABLE 0x0001
+#define SEEN      0x0002
+
+static int show_root;
+static int show_tags;
+static int show_unreachable;
+static int check_full;
+static int check_strict;
+static int keep_cache_objects;
+static unsigned char head_sha1[20];
+
+#ifdef NO_D_INO_IN_DIRENT
+#define SORT_DIRENT 0
+#define DIRENT_SORT_HINT(de) 0
+#else
+#define SORT_DIRENT 1
+#define DIRENT_SORT_HINT(de) ((de)->d_ino)
+#endif
+
+static void objreport(struct object *obj, const char *severity,
+                      const char *err, va_list params)
+{
+       fprintf(stderr, "%s in %s %s: ",
+               severity, typename(obj->type), sha1_to_hex(obj->sha1));
+       vfprintf(stderr, err, params);
+       fputs("\n", stderr);
+}
+
+static int objerror(struct object *obj, const char *err, ...)
+{
+       va_list params;
+       va_start(params, err);
+       objreport(obj, "error", err, params);
+       va_end(params);
+       return -1;
+}
+
+static int objwarning(struct object *obj, const char *err, ...)
+{
+       va_list params;
+       va_start(params, err);
+       objreport(obj, "warning", err, params);
+       va_end(params);
+       return -1;
+}
+
+/*
+ * Check a single reachable object
+ */
+static void check_reachable_object(struct object *obj)
+{
+       const struct object_refs *refs;
+
+       /*
+        * We obviously want the object to be parsed,
+        * except if it was in a pack-file and we didn't
+        * do a full fsck
+        */
+       if (!obj->parsed) {
+               if (has_sha1_file(obj->sha1))
+                       return; /* it is in pack - forget about it */
+               printf("missing %s %s\n", typename(obj->type), sha1_to_hex(obj->sha1));
+               return;
+       }
+
+       /*
+        * Check that everything that we try to reference is also good.
+        */
+       refs = lookup_object_refs(obj);
+       if (refs) {
+               unsigned j;
+               for (j = 0; j < refs->count; j++) {
+                       struct object *ref = refs->ref[j];
+                       if (ref->parsed ||
+                           (has_sha1_file(ref->sha1)))
+                               continue;
+                       printf("broken link from %7s %s\n",
+                              typename(obj->type), sha1_to_hex(obj->sha1));
+                       printf("              to %7s %s\n",
+                              typename(ref->type), sha1_to_hex(ref->sha1));
+               }
+       }
+}
+
+/*
+ * Check a single unreachable object
+ */
+static void check_unreachable_object(struct object *obj)
+{
+       /*
+        * Missing unreachable object? Ignore it. It's not like
+        * we miss it (since it can't be reached), nor do we want
+        * to complain about it being unreachable (since it does
+        * not exist).
+        */
+       if (!obj->parsed)
+               return;
+
+       /*
+        * Unreachable object that exists? Show it if asked to,
+        * since this is something that is prunable.
+        */
+       if (show_unreachable) {
+               printf("unreachable %s %s\n", typename(obj->type), sha1_to_hex(obj->sha1));
+               return;
+       }
+
+       /*
+        * "!used" means that nothing at all points to it, including
+        * other unreacahble objects. In other words, it's the "tip"
+        * of some set of unreachable objects, usually a commit that
+        * got dropped.
+        *
+        * Such starting points are more interesting than some random
+        * set of unreachable objects, so we show them even if the user
+        * hasn't asked for _all_ unreachable objects. If you have
+        * deleted a branch by mistake, this is a prime candidate to
+        * start looking at, for example.
+        */
+       if (!obj->used) {
+               printf("dangling %s %s\n", typename(obj->type),
+                      sha1_to_hex(obj->sha1));
+               return;
+       }
+
+       /*
+        * Otherwise? It's there, it's unreachable, and some other unreachable
+        * object points to it. Ignore it - it's not interesting, and we showed
+        * all the interesting cases above.
+        */
+}
+
+static void check_object(struct object *obj)
+{
+       if (obj->flags & REACHABLE)
+               check_reachable_object(obj);
+       else
+               check_unreachable_object(obj);
+}
+
+static void check_connectivity(void)
+{
+       int i, max;
+
+       /* Look up all the requirements, warn about missing objects.. */
+       max = get_max_object_index();
+       for (i = 0; i < max; i++) {
+               struct object *obj = get_indexed_object(i);
+
+               if (obj)
+                       check_object(obj);
+       }
+}
+
+/*
+ * The entries in a tree are ordered in the _path_ order,
+ * which means that a directory entry is ordered by adding
+ * a slash to the end of it.
+ *
+ * So a directory called "a" is ordered _after_ a file
+ * called "a.c", because "a/" sorts after "a.c".
+ */
+#define TREE_UNORDERED (-1)
+#define TREE_HAS_DUPS  (-2)
+
+static int verify_ordered(unsigned mode1, const char *name1, unsigned mode2, const char *name2)
+{
+       int len1 = strlen(name1);
+       int len2 = strlen(name2);
+       int len = len1 < len2 ? len1 : len2;
+       unsigned char c1, c2;
+       int cmp;
+
+       cmp = memcmp(name1, name2, len);
+       if (cmp < 0)
+               return 0;
+       if (cmp > 0)
+               return TREE_UNORDERED;
+
+       /*
+        * Ok, the first <len> characters are the same.
+        * Now we need to order the next one, but turn
+        * a '\0' into a '/' for a directory entry.
+        */
+       c1 = name1[len];
+       c2 = name2[len];
+       if (!c1 && !c2)
+               /*
+                * git-write-tree used to write out a nonsense tree that has
+                * entries with the same name, one blob and one tree.  Make
+                * sure we do not have duplicate entries.
+                */
+               return TREE_HAS_DUPS;
+       if (!c1 && S_ISDIR(mode1))
+               c1 = '/';
+       if (!c2 && S_ISDIR(mode2))
+               c2 = '/';
+       return c1 < c2 ? 0 : TREE_UNORDERED;
+}
+
+static int fsck_tree(struct tree *item)
+{
+       int retval;
+       int has_full_path = 0;
+       int has_zero_pad = 0;
+       int has_bad_modes = 0;
+       int has_dup_entries = 0;
+       int not_properly_sorted = 0;
+       struct tree_desc desc;
+       unsigned o_mode;
+       const char *o_name;
+       const unsigned char *o_sha1;
+
+       desc.buf = item->buffer;
+       desc.size = item->size;
+
+       o_mode = 0;
+       o_name = NULL;
+       o_sha1 = NULL;
+       while (desc.size) {
+               unsigned mode;
+               const char *name;
+               const unsigned char *sha1;
+
+               sha1 = tree_entry_extract(&desc, &name, &mode);
+
+               if (strchr(name, '/'))
+                       has_full_path = 1;
+               has_zero_pad |= *(char *)desc.buf == '0';
+               update_tree_entry(&desc);
+
+               switch (mode) {
+               /*
+                * Standard modes..
+                */
+               case S_IFREG | 0755:
+               case S_IFREG | 0644:
+               case S_IFLNK:
+               case S_IFDIR:
+                       break;
+               /*
+                * This is nonstandard, but we had a few of these
+                * early on when we honored the full set of mode
+                * bits..
+                */
+               case S_IFREG | 0664:
+                       if (!check_strict)
+                               break;
+               default:
+                       has_bad_modes = 1;
+               }
+
+               if (o_name) {
+                       switch (verify_ordered(o_mode, o_name, mode, name)) {
+                       case TREE_UNORDERED:
+                               not_properly_sorted = 1;
+                               break;
+                       case TREE_HAS_DUPS:
+                               has_dup_entries = 1;
+                               break;
+                       default:
+                               break;
+                       }
+               }
+
+               o_mode = mode;
+               o_name = name;
+               o_sha1 = sha1;
+       }
+       free(item->buffer);
+       item->buffer = NULL;
+
+       retval = 0;
+       if (has_full_path) {
+               objwarning(&item->object, "contains full pathnames");
+       }
+       if (has_zero_pad) {
+               objwarning(&item->object, "contains zero-padded file modes");
+       }
+       if (has_bad_modes) {
+               objwarning(&item->object, "contains bad file modes");
+       }
+       if (has_dup_entries) {
+               retval = objerror(&item->object, "contains duplicate file entries");
+       }
+       if (not_properly_sorted) {
+               retval = objerror(&item->object, "not properly sorted");
+       }
+       return retval;
+}
+
+static int fsck_commit(struct commit *commit)
+{
+       char *buffer = commit->buffer;
+       unsigned char tree_sha1[20], sha1[20];
+
+       if (memcmp(buffer, "tree ", 5))
+               return objerror(&commit->object, "invalid format - expected 'tree' line");
+       if (get_sha1_hex(buffer+5, tree_sha1) || buffer[45] != '\n')
+               return objerror(&commit->object, "invalid 'tree' line format - bad sha1");
+       buffer += 46;
+       while (!memcmp(buffer, "parent ", 7)) {
+               if (get_sha1_hex(buffer+7, sha1) || buffer[47] != '\n')
+                       return objerror(&commit->object, "invalid 'parent' line format - bad sha1");
+               buffer += 48;
+       }
+       if (memcmp(buffer, "author ", 7))
+               return objerror(&commit->object, "invalid format - expected 'author' line");
+       free(commit->buffer);
+       commit->buffer = NULL;
+       if (!commit->tree)
+               return objerror(&commit->object, "could not load commit's tree %s", tree_sha1);
+       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_tag(struct tag *tag)
+{
+       struct object *tagged = tag->tagged;
+
+       if (!tagged) {
+               return objerror(&tag->object, "could not load tagged object");
+       }
+       if (!show_tags)
+               return 0;
+
+       printf("tagged %s %s", typename(tagged->type), sha1_to_hex(tagged->sha1));
+       printf(" (%s) in %s\n", tag->tag, sha1_to_hex(tag->object.sha1));
+       return 0;
+}
+
+static int fsck_sha1(unsigned char *sha1)
+{
+       struct object *obj = parse_object(sha1);
+       if (!obj)
+               return error("%s: object corrupt or missing", sha1_to_hex(sha1));
+       if (obj->flags & SEEN)
+               return 0;
+       obj->flags |= SEEN;
+       if (obj->type == OBJ_BLOB)
+               return 0;
+       if (obj->type == OBJ_TREE)
+               return fsck_tree((struct tree *) obj);
+       if (obj->type == OBJ_COMMIT)
+               return fsck_commit((struct commit *) obj);
+       if (obj->type == OBJ_TAG)
+               return fsck_tag((struct tag *) obj);
+       /* By now, parse_object() would've returned NULL instead. */
+       return objerror(obj, "unknown type '%d' (internal fsck error)", obj->type);
+}
+
+/*
+ * 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];
+};
+
+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;
+
+       if (SORT_DIRENT)
+               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;
+               fsck_sha1(sha1);
+               free(entry);
+       }
+       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;
+       hashcpy(entry->sha1, sha1);
+       nr = sha1_list.nr;
+       if (nr == MAX_SHA1_ENTRIES) {
+               fsck_sha1_list();
+               nr = 0;
+       }
+       sha1_list.entry[nr] = entry;
+       sha1_list.nr = ++nr;
+}
+
+static void fsck_dir(int i, char *path)
+{
+       DIR *dir = opendir(path);
+       struct dirent *de;
+
+       if (!dir)
+               return;
+
+       while ((de = readdir(dir)) != NULL) {
+               char name[100];
+               unsigned char sha1[20];
+               int len = strlen(de->d_name);
+
+               switch (len) {
+               case 2:
+                       if (de->d_name[1] != '.')
+                               break;
+               case 1:
+                       if (de->d_name[0] != '.')
+                               break;
+                       continue;
+               case 38:
+                       sprintf(name, "%02x", i);
+                       memcpy(name+2, de->d_name, len+1);
+                       if (get_sha1_hex(name, sha1) < 0)
+                               break;
+                       add_sha1_list(sha1, DIRENT_SORT_HINT(de));
+                       continue;
+               }
+               fprintf(stderr, "bad sha1 file: %s/%s\n", path, de->d_name);
+       }
+       closedir(dir);
+}
+
+static int default_refs;
+
+static int fsck_handle_reflog_ent(unsigned char *osha1, unsigned char *nsha1,
+               const char *email, unsigned long timestamp, int tz,
+               const char *message, void *cb_data)
+{
+       struct object *obj;
+
+       if (!is_null_sha1(osha1)) {
+               obj = lookup_object(osha1);
+               if (obj) {
+                       obj->used = 1;
+                       mark_reachable(obj, REACHABLE);
+               }
+       }
+       obj = lookup_object(nsha1);
+       if (obj) {
+               obj->used = 1;
+               mark_reachable(obj, REACHABLE);
+       }
+       return 0;
+}
+
+static int fsck_handle_ref(const char *refname, const unsigned char *sha1, int flag, void *cb_data)
+{
+       struct object *obj;
+
+       obj = lookup_object(sha1);
+       if (!obj) {
+               if (has_sha1_file(sha1)) {
+                       default_refs++;
+                       return 0; /* it is in a pack */
+               }
+               error("%s: invalid sha1 pointer %s", refname, sha1_to_hex(sha1));
+               /* We'll continue with the rest despite the error.. */
+               return 0;
+       }
+       default_refs++;
+       obj->used = 1;
+       mark_reachable(obj, REACHABLE);
+
+       for_each_reflog_ent(refname, fsck_handle_reflog_ent, NULL);
+
+       return 0;
+}
+
+static void get_default_heads(void)
+{
+       for_each_ref(fsck_handle_ref, NULL);
+
+       /*
+        * Not having any default heads isn't really fatal, but
+        * it does mean that "--unreachable" no longer makes any
+        * sense (since in this case everything will obviously
+        * be unreachable by definition.
+        *
+        * Showing dangling objects is valid, though (as those
+        * dangling objects are likely lost heads).
+        *
+        * So we just print a warning about it, and clear the
+        * "show_unreachable" flag.
+        */
+       if (!default_refs) {
+               error("No default references");
+               show_unreachable = 0;
+       }
+}
+
+static void fsck_object_dir(const char *path)
+{
+       int i;
+       for (i = 0; i < 256; i++) {
+               static char dir[4096];
+               sprintf(dir, "%s/%02x", path, i);
+               fsck_dir(i, dir);
+       }
+       fsck_sha1_list();
+}
+
+static int fsck_head_link(void)
+{
+       unsigned char sha1[20];
+       int flag;
+       const char *head_points_at = resolve_ref("HEAD", sha1, 1, &flag);
+
+       if (!head_points_at || !(flag & REF_ISSYMREF))
+               return error("HEAD is not a symbolic ref");
+       if (strncmp(head_points_at, "refs/heads/", 11))
+               return error("HEAD points to something strange (%s)",
+                            head_points_at);
+       if (is_null_sha1(sha1))
+               return error("HEAD: not a valid git pointer");
+       return 0;
+}
+
+static int fsck_cache_tree(struct cache_tree *it)
+{
+       int i;
+       int err = 0;
+
+       if (0 <= it->entry_count) {
+               struct object *obj = parse_object(it->sha1);
+               if (!obj) {
+                       error("%s: invalid sha1 pointer in cache-tree",
+                             sha1_to_hex(it->sha1));
+                       return 1;
+               }
+               mark_reachable(obj, REACHABLE);
+               obj->used = 1;
+               if (obj->type != OBJ_TREE)
+                       err |= objerror(obj, "non-tree in cache-tree");
+       }
+       for (i = 0; i < it->subtree_nr; i++)
+               err |= fsck_cache_tree(it->down[i]->cache_tree);
+       return err;
+}
+
+int main(int argc, char **argv)
+{
+       int i, heads;
+
+       track_object_refs = 1;
+       setup_git_directory();
+
+       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 (!strcmp(arg, "--cache")) {
+                       keep_cache_objects = 1;
+                       continue;
+               }
+               if (!strcmp(arg, "--full")) {
+                       check_full = 1;
+                       continue;
+               }
+               if (!strcmp(arg, "--strict")) {
+                       check_strict = 1;
+                       continue;
+               }
+               if (*arg == '-')
+                       usage("git-fsck [--tags] [--root] [[--unreachable] [--cache] [--full] [--strict] <head-sha1>*]");
+       }
+
+       fsck_head_link();
+       fsck_object_dir(get_object_directory());
+       if (check_full) {
+               struct alternate_object_database *alt;
+               struct packed_git *p;
+               prepare_alt_odb();
+               for (alt = alt_odb_list; alt; alt = alt->next) {
+                       char namebuf[PATH_MAX];
+                       int namelen = alt->name - alt->base;
+                       memcpy(namebuf, alt->base, namelen);
+                       namebuf[namelen - 1] = 0;
+                       fsck_object_dir(namebuf);
+               }
+               prepare_packed_git();
+               for (p = packed_git; p; p = p->next)
+                       /* verify gives error messages itself */
+                       verify_pack(p, 0);
+
+               for (p = packed_git; p; p = p->next) {
+                       int num = num_packed_objects(p);
+                       for (i = 0; i < num; i++) {
+                               unsigned char sha1[20];
+                               nth_packed_object_sha1(p, i, sha1);
+                               fsck_sha1(sha1);
+                       }
+               }
+       }
+
+       heads = 0;
+       for (i = 1; i < argc; i++) {
+               const char *arg = argv[i]; 
+
+               if (*arg == '-')
+                       continue;
+
+               if (!get_sha1(arg, head_sha1)) {
+                       struct object *obj = lookup_object(head_sha1);
+
+                       /* Error is printed by lookup_object(). */
+                       if (!obj)
+                               continue;
+
+                       obj->used = 1;
+                       mark_reachable(obj, REACHABLE);
+                       heads++;
+                       continue;
+               }
+               error("invalid parameter: expected sha1, got '%s'", arg);
+       }
+
+       /*
+        * If we've not been given any explicit head information, do the
+        * default ones from .git/refs. We also consider the index file
+        * in this case (ie this implies --cache).
+        */
+       if (!heads) {
+               get_default_heads();
+               keep_cache_objects = 1;
+       }
+
+       if (keep_cache_objects) {
+               int i;
+               read_cache();
+               for (i = 0; i < active_nr; i++) {
+                       struct blob *blob = lookup_blob(active_cache[i]->sha1);
+                       struct object *obj;
+                       if (!blob)
+                               continue;
+                       obj = &blob->object;
+                       obj->used = 1;
+                       mark_reachable(obj, REACHABLE);
+               }
+               if (active_cache_tree)
+                       fsck_cache_tree(active_cache_tree);
+       }
+
+       check_connectivity();
+       return 0;
+}
index b928f2ca52c1991ee50838fc47580ed83c58fb64..9360804711d6b81983e53df67c86ddc8082e2ab4 100755 (executable)
@@ -12,7 +12,7 @@ fi
 laf="$GIT_DIR/lost-found"
 rm -fr "$laf" && mkdir -p "$laf/commit" "$laf/other" || exit
 
-git fsck-objects --full |
+git fsck --full |
 while read dangling type sha1
 do
        case "$dangling" in
index efd494a47fb961f4f8425d52bd10e37e5ffdab51..9f790a08e5d2760524ab7a34e98071062b3ed42f 100644 (file)
@@ -1069,7 +1069,7 @@ int main(int argc, const char **argv)
                fprintf(stderr,
 "Some loose object were found to be corrupt, but they might be just\n"
 "a false '404 Not Found' error message sent with incorrect HTTP\n"
-"status code.  Suggest running git fsck-objects.\n");
+"status code.  Suggest running git-fsck.\n");
        }
        return rc;
 }
index 47d124732cd04481d6a21e927aa6104f63e21fbf..e5bbc384f7f4e56b13a13bf2bedad1427e257652 100755 (executable)
@@ -20,7 +20,7 @@ check_have () {
 }
 
 check_fsck () {
-       output=$(git fsck-objects --full)
+       output=$(git fsck --full)
        case "$1" in
        '')
                test -z "$output" ;;
index e35d60fa3ab9d7694f89f4451caf741003feeb13..48e3d1705f5e4bc7f206692276b4e3e1fe1ddf66 100755 (executable)
@@ -63,13 +63,13 @@ pull_to_client () {
        case "$heads" in *B*) echo $BTIP > .git/refs/heads/B;; esac
        git-symbolic-ref HEAD refs/heads/`echo $heads | sed -e 's/^\(.\).*$/\1/'`
 
-       test_expect_success "fsck" 'git-fsck-objects --full > fsck.txt 2>&1'
+       test_expect_success "fsck" 'git-fsck --full > fsck.txt 2>&1'
 
        test_expect_success 'check downloaded results' \
        'mv .git/objects/pack/pack-* . &&
         p=`ls -1 pack-*.pack` &&
         git-unpack-objects <$p &&
-        git-fsck-objects --full'
+        git-fsck --full'
 
        test_expect_success "new object count after $number pull" \
        'idx=`echo pack-*.idx` &&
@@ -145,7 +145,7 @@ test_expect_success "clone shallow object count (part 2)" '
 '
 
 test_expect_success "fsck in shallow repo" \
-       "(cd shallow; git-fsck-objects --full)"
+       "(cd shallow; git-fsck --full)"
 
 #test_done; exit
 
index b9f6d9636313c29cb9d9553ca908e623fdefd179..2f8e97cb7ea4b8bdf56d5795d512534d504d25c0 100755 (executable)
@@ -17,7 +17,7 @@ reachable_via() {
 }
 
 test_valid_repo() {
-       git fsck-objects --full > fsck.log &&
+       git fsck --full > fsck.log &&
        test `wc -l < fsck.log` = 0
 }