Re: fetch failure from git2.kernel.org?

[Junio C Hamano <junkio@cox.net>]

Linus Torvalds <torvalds@linux-foundation.org> writes:

> On Sun, 22 Apr 2007, Randal L. Schwartz wrote:
>> 
>> is this known?  transient?
>
> Junio seems to have tightened his permissions on kernel.org, and as a 
> result the most recent pack is unreadable by anybody but him (I'm not sure 
> if such private files even get mirrored out - but even if they do, they 
> will get mirrored out with the same tight permissions and won't be 
> readable).

That's not *me* per-se, it's Nico ;-). 

Anybody who runs the new pack-objects that has been cooking in
'next' would get this.

I think this patch minimally fixes it, but some existing tests
(e.g. t5300) assume that pack/idx are writable which makes them
break.


diff --git a/builtin-pack-objects.c b/builtin-pack-objects.c
index c72e07a..4eb88e8 100644
--- a/builtin-pack-objects.c
+++ b/builtin-pack-objects.c
@@ -1783,11 +1783,17 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
 		write_index_file(last_obj_offset, object_list_sha1);
 		snprintf(tmpname, sizeof(tmpname), "%s-%s.pack",
 			 base_name, sha1_to_hex(object_list_sha1));
+		if (chmod(pack_tmp_name, 0444))
+			die("unable to make temporary pack file readable: %s",
+			    strerror(errno));
 		if (rename(pack_tmp_name, tmpname))
 			die("unable to rename temporary pack file: %s",
 			    strerror(errno));
 		snprintf(tmpname, sizeof(tmpname), "%s-%s.idx",
 			 base_name, sha1_to_hex(object_list_sha1));
+		if (chmod(idx_tmp_name, 0444))
+			die("unable to make temporary index file readable: %s",
+			    strerror(errno));
 		if (rename(idx_tmp_name, tmpname))
 			die("unable to rename temporary index file: %s",
 			    strerror(errno));