Jeff King wrote:
> Commit b2f2039c2b (fsck: accept an oid instead of a "struct tree" for
> fsck_tree(), 2019-10-18) introduced a new "oid" parameter to
> fsck_tree(), and we pass it to the report() function when we find
> problems. However, that is shadowed within the tree-walking loop by the
> existing "oid" variable which we use to store the oid of each tree
> entry. As a result, we may report the wrong oid for some problems we
> detect within the loop (the entry oid, instead of the tree oid).
>
> Our tests didn't catch this because they checked only that we found the
> expected fsck problem, not that it was attached to the correct object.
Oh, goodness. Does this mean we should be similarly checking oid in
the rest of the fsck test scripts? (I'm not saying this patch should
do so, just curious about what you think on the subject.)
> Let's rename both variables in the function to avoid confusion. This
> makes the diff a little noisy (e.g., all of the report() calls outside
> the loop wee already correct but need touched), but makes sure we catch
nit: s/wee/are/, s/need touched/need to be touched/
> all cases and will avoid similar confusion in the future.
Thanks for doing that --- it does make reviewing easier.
> Signed-off-by: Jeff King <peff@xxxxxxxx>
> ---
> fsck.c | 40 +++++++++++++++++++-------------------
> t/t7415-submodule-names.sh | 5 +++--
> 2 files changed, 23 insertions(+), 22 deletions(-)
>
> diff --git a/fsck.c b/fsck.c
> index f82e2fe9e3..46a108839f 100644
> --- a/fsck.c
> +++ b/fsck.c
> @@ -633,7 +633,7 @@ static int verify_ordered(unsigned mode1, const char *name1,
> return c1 < c2 ? 0 : TREE_UNORDERED;
> }
>
> -static int fsck_tree(const struct object_id *oid,
> +static int fsck_tree(const struct object_id *tree_oid,
optional: we could call it "tree".
> const char *buffer, unsigned long size,
> struct fsck_options *options)
> {
> @@ -654,7 +654,7 @@ static int fsck_tree(const struct object_id *oid,
> struct name_stack df_dup_candidates = { NULL };
>
> if (init_tree_desc_gently(&desc, buffer, size)) {
> - retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
> + retval += report(options, tree_oid, OBJ_TREE, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
> return retval;
> }
>
> @@ -664,11 +664,11 @@ static int fsck_tree(const struct object_id *oid,
> while (desc.size) {
> unsigned short mode;
> const char *name, *backslash;
> - const struct object_id *oid;
> + const struct object_id *entry_oid;
>
> - oid = tree_entry_extract(&desc, &name, &mode);
> + entry_oid = tree_entry_extract(&desc, &name, &mode);
optional: could call it "child".
>
> - has_null_sha1 |= is_null_oid(oid);
> + has_null_sha1 |= is_null_oid(entry_oid);
> has_full_path |= !!strchr(name, '/');
> has_empty_name |= !*name;
> has_dot |= !strcmp(name, ".");
> @@ -678,10 +678,10 @@ static int fsck_tree(const struct object_id *oid,
>
> if (is_hfs_dotgitmodules(name) || is_ntfs_dotgitmodules(name)) {
> if (!S_ISLNK(mode))
> - oidset_insert(&gitmodules_found, oid);
> + oidset_insert(&gitmodules_found, entry_oid);
> else
> retval += report(options,
> - oid, OBJ_TREE,
> + tree_oid, OBJ_TREE,
> FSCK_MSG_GITMODULES_SYMLINK,
> ".gitmodules is a symbolic link");
Right, this is a property of the enclosing tree, not the blob
representing the symlink target. Good.
not about this patch: Could report() notice when the oid doesn't match
the type passed in, to more easily catch this kind of mistake?
[...]
> @@ -692,9 +692,9 @@ static int fsck_tree(const struct object_id *oid,
> has_dotgit |= is_ntfs_dotgit(backslash);
> if (is_ntfs_dotgitmodules(backslash)) {
> if (!S_ISLNK(mode))
> - oidset_insert(&gitmodules_found, oid);
> + oidset_insert(&gitmodules_found, entry_oid);
> else
> - retval += report(options, oid, OBJ_TREE,
> + retval += report(options, tree_oid, OBJ_TREE,
> FSCK_MSG_GITMODULES_SYMLINK,
> ".gitmodules is a symbolic link");
Likewise, good.
[...]
> --- a/t/t7415-submodule-names.sh
> +++ b/t/t7415-submodule-names.sh
> @@ -148,13 +148,14 @@ test_expect_success 'fsck detects symlinked .gitmodules file' '
> {
> printf "100644 blob $content\t$tricky\n" &&
> printf "120000 blob $target\t.gitmodules\n"
> - } | git mktree &&
> + } >bad-tree &&
> + tree=$(git mktree <bad-tree) &&
>
> # Check not only that we fail, but that it is due to the
> # symlink detector; this grep string comes from the config
> # variable name and will not be translated.
> test_must_fail git fsck 2>output &&
> - test_i18ngrep gitmodulesSymlink output
> + test_i18ngrep "tree $tree: gitmodulesSymlink" output
Makes sense.
By the way, why does GETTEXT_POISON lose the gitmodulesSymlink
keyword? Is this just a limitation of GETTEXT_POISON losing
information that's passed in with %s?
With the commit message tweak mentioned above,
Reviewed-by: Jonathan Nieder <jrnieder@xxxxxxxxx>
Thanks.
