This means that
  git show HEAD:
will now return HEAD^{tree}, which is logically consistent with
  git show HEAD:Documentation

Signed-off-by: Jeff King <peff@peff.net>
---
 tree-walk.c |    9 ++++++++-
 1 files changed, 8 insertions(+), 1 deletions(-)

diff --git a/tree-walk.c b/tree-walk.c
index 22f4550..70f8999 100644
--- a/tree-walk.c
+++ b/tree-walk.c
@@ -199,10 +199,17 @@ int get_tree_entry(const unsigned char *tree_sha1, const char *name, unsigned ch
 	int retval;
 	void *tree;
 	struct tree_desc t;
+	unsigned char root[20];
 
-	tree = read_object_with_reference(tree_sha1, tree_type, &t.size, NULL);
+	tree = read_object_with_reference(tree_sha1, tree_type, &t.size, root);
 	if (!tree)
 		return -1;
+
+	if (name[0] == '\0') {
+		hashcpy(sha1, root);
+		return 0;
+	}
+
 	t.buf = tree;
 	retval = find_tree_entry(&t, name, sha1, mode);
 	free(tree);
-- 
1.4.4.4.g9612-dirty
-

[ message continues ]

Hi,


Good idea!

Ciao,
Dscho

-

[ message continues ]

Makes sense, although I somewhat worry about changing the
behaviour and potentially breaking the expectation of callers
of get_tree_entry() other than "git show".


-

[ message continues ]

Previously, trying to get an empty name would always return -1. Most of
the users besides sha1_name are getting the paths by enumerating the
index or an existing tree, so they will never be blank. The exceptions
are:
 - git-blame ''; this fails in both the root and a subdir (though with
   different errors); behavior is unchanged with my patch
 - git-archive feeds the 'prefix' variable to get_tree_entry; it works
   fine in the root with or without the patch. I seem to be getting an
   error, with or without my patch, when doing a git-archive from a
   subdirectory:
     $ git-archive --format=tar origin >/dev/null
     $ cd Documentation
     $ git-archive --format=tar origin >/dev/null
     fatal: cannot read c87c61af00c6d2cd7212240e260809000000aaab
   but I haven't been able to track it down further. It's clearly
   unrelated to my patch, though.

The main difference is that looking up a ref through sha1_name used to
fail to match 'HEAD:'; now it does. So if you were trying to match a
file 'HEAD:' without using '--', you're out of luck.

-Peff
-

[ message continues ]

I've noticed this while looking at your patch and fixed it in my
tree already.

-

[ message continues ]

Thanks, I will stop looking, then. :)

-Peff
-

[ message continues ]

-- >8 --
[PATCH] builtin-archive: do not free a tree held by the object layer.

Found by running "git archive --format=tar HEAD" in Documentation/
directory.

It's surprising that nobody has noticed this from the beginning...

Signed-off-by: Junio C Hamano <junkio@cox.net>
---
 builtin-archive.c |    1 -
 1 files changed, 0 insertions(+), 1 deletions(-)

diff --git a/builtin-archive.c b/builtin-archive.c
index 391cf43..32737d3 100644
--- a/builtin-archive.c
+++ b/builtin-archive.c
@@ -137,7 +137,6 @@ void parse_treeish_arg(const char **argv, struct archiver_args *ar_args,
 		if (err || !S_ISDIR(mode))
 			die("current working directory is untracked");
 
-		free(tree);
 		tree = parse_tree_indirect(tree_sha1);
 	}
 	ar_args->tree = tree;

-

[ message continues ]