Re: [PATCH] git-svn: fix ls-tree usage with dash-prefixed paths

From: Junio C Hamano <gitster@pobox.com>
Date: Mar 29, 2009, 20:33 UTC
Message-ID: <7v8wmoqdc1.fsf@gitster.siamese.dyndns.org>
In-Reply-To: <20090329061045.GA29721@dcvr.yhbt.net>

Eric Wong <normalperson@yhbt.net> writes:

Show 13 quoted lines

> To find the blob object name given a tree and pathname, we were
> incorrectly calling "git ls-tree" with a "--" argument followed
> by the pathname of the file we wanted to get.
>
>   git ls-tree <TREE> -- --dashed/path/name.c
>
> Unlike many command-line interfaces, the "--" alone does not
> symbolize the end of non-option arguments on the command-line.
>
> ls-tree interprets the "--" as a prefix to match against, thus
> the entire contents of the --dashed/* hierarchy would be
> returned because the "--" matches "--dashed" and every path
> under it.

The above makes only half a sense to me.  In an empty directory:

    $ git init
    Initialized empty Git repository in /tmp/empty/.git
    $ mkdir -p ./--dashed/path
    $ >./--dashed/path/name
    $ git add .
    $ git ls-files
    --dashed/path/name
    $ git commit -a -m initial
    [master (root-commit) cd44284] initial
     0 files changed, 0 insertions(+), 0 deletions(-)
     create mode 100644 --dashed/path/name
    $ git ls-tree HEAD^{tree} --
    $ git ls-tree HEAD^{tree} -- --dashed/path/name
    100644 blob e69de29bb2d1d6434b8b29ae775ad8c2e48c5391	--dashed/path/name
    $ mkdir ./--
    $ >./--/eman
    $ git add .
    $ git commit -m second
    [master 80f8ef9] second
     0 files changed, 0 insertions(+), 0 deletions(-)
     create mode 100644 --/eman
    $ git ls-tree HEAD^{tree} -- --dashed/path
    100644 blob e69de29bb2d1d6434b8b29ae775ad8c2e48c5391	--/eman
    040000 tree 23e59e0c91294c39ac7c5a2e39efb01d878de9a0	--dashed/path
    $ exit

Perhaps the problem repository had a pathname that is exactly -- (in addition to --dashed/), and ls-tree emitted everything under --/ hierarchy? In other words, your fix to git-svn may be correct and I am reading your problem description above incorrectly?

As the command always takes exactly one tree, it could be argued that it is not a bug that it does not honour the usual -- convention, even though I am tempted to think it is of a very dark shade of gray. It is certainly something that we would have done differently if we were implementing the command today.

"Fixing" ls-tree would be trivial to ignore the first "--" if it precedes other pathspecs (see below), but the command is a plumbing, and such a change will break existing scripts that have relied on the existing behaviour since 2005, so I do not think it is worth the risk of causing such silent breakages to them. Besides, with such a "fix", fixing of user scripts will become much more cumbersome, as they need to detect the version of git and drive ls-tree differently.

 builtin-ls-tree.c |    6 ++++++
 1 files changed, 6 insertions(+), 0 deletions(-)

diff --git a/builtin-ls-tree.c b/builtin-ls-tree.c
index 22008df..08c4307 100644
--- a/builtin-ls-tree.c
+++ b/builtin-ls-tree.c
@@ -186,6 +186,12 @@ int cmd_ls_tree(int argc, const char **argv, const char *prefix)
 	if (get_sha1(argv[1], sha1))
 		die("Not a valid object name %s", argv[1]);
 
+	if (3 < argc && !strcmp(argv[2], "--")) {
+		/* ls-tree <tree> -- pathspec */
+		argc--;
+		argv++;
+		warning("ignoring -- in 'ls-tree <tree> -- <pathspec>'");
+	}
 	pathspec = get_pathspec(prefix, argv + 2);
 	tree = parse_tree_indirect(sha1);
 	if (!tree)