Re: error pushing stash ?

[Jeff King <peff@peff.net>]

On Mon, Oct 06, 2008 at 05:40:51PM -0700, Shawn O. Pearce wrote:

> >  ! [remote rejected] refs/stash -> refs/stash (funny refname)
> > error: failed to push some refs to '/users/dbryson/backup/janus.git/'
> 
> refs/stash is a funny refname because it contains only 1 '/'.
> Normally a valid ref has at least 2 '/', e.g. refs/heads/8654 or
> refs/tags/v1.0.

Since no version of receive-pack accepts these "funny refs", perhaps we
should mirror the check when considering the list of refs to send. IOW,
don't even make them eligible for matching or mirroring. Patch is below.

> Naming the stash refs/stash was perhaps funny in the first place
> since it cannot be moved about on the transport protocol, but then
> again the bulk of the stash data is actually in the reflog for the
> stash (and not the stash ref itself) so there is basically no point
> in pushing or fetching a stash directly.

I agree there is not much point in pushing it, since the useful bit is
in the reflog. So perhaps a "funny" refname is a good place to put it,
since it easily tells us that it is not a useful thing to push.

---
diff --git a/builtin-send-pack.c b/builtin-send-pack.c
index bbf6e0a..298bd71 100644
--- a/builtin-send-pack.c
+++ b/builtin-send-pack.c
@@ -140,7 +140,13 @@ static struct ref *remote_refs, **remote_tail;
 static int one_local_ref(const char *refname, const unsigned char *sha1, int flag, void *cb_data)
 {
 	struct ref *ref;
-	int len = strlen(refname) + 1;
+	int len;
+
+	/* we already know it starts with refs/ to get here */
+	if (check_ref_format(refname + 5))
+		return 0;
+
+	len = strlen(refname) + 1;
 	ref = xcalloc(1, sizeof(*ref) + len);
 	hashcpy(ref->new_sha1, sha1);
 	memcpy(ref->name, refname, len);