Re: What's cooking in git/spearce.git (Oct 2008, #02; Sun, 12)

Junio C Hamano [off-list ref] writes:

"Shawn O. Pearce" [off-list ref] writes:

quoted

* pb/rename-rowin32 (Fri Oct 3 12:20:43 2008 +0200) 1 commit
 - Do not rename read-only files during a push

Supposedly fixes pack file renames on Windows, but it makes the
test suite fail on Linux.  I haven't debugged why yet.

I am kind of surprised that it actually passes the test on Windows, which
implies that these cats shown in the patch to fix this breakage below do
not honor ro-ness of the file, which in turn makes me doubt if making the
resulting packfiles read-only has any effect on that platform whatsoever..

Actually, the patch I posted was wrong and shouldn't be applied, as the
original breakage is a sign that the externally observable behaviour of
the command has changed, and the patch was just hiding it under the rug.

The thing is that 8c76006 (Do not rename read-only files during a push,
2008-10-03) makes unnecessary chmod() when there is no rename is involved,
namely when it is used to index an existing packfile on the filesystem.

I think the attached patch is a better fix, to be squashed in.

Note: I now have the pumpkin, so you do not have to apply and push the
results out --- this is just me showing a proposed change as usual.

 index-pack.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git c/index-pack.c w/index-pack.c
index 2c69f08..aec11cb 100644
--- c/index-pack.c
+++ w/index-pack.c

@@ -823,7 +823,8 @@ static void final(const char *final_pack_name, const char *curr_pack_name,
 		if (move_temp_to_file(curr_pack_name, final_pack_name))
 			die("cannot store pack file");
 	}
-	chmod(final_pack_name, 0444);
+	if (from_stdin)
+		chmod(final_pack_name, 0444);
 
 	if (final_index_name != curr_index_name) {
 		if (!final_index_name) {