fetch refspec foo/* matches foo*
- From
Jeff King <peff@peff.net>
- Date
- Jul 26, 2008, 08:24 UTC
- Message-ID
- <20080726082405.GA10104@sigill.intra.peff.net>
- In-Reply-To
- <7v1w1hsmnc.fsf@gitster.siamese.dyndns.org>
On Fri, Jul 25, 2008 at 02:02:15PM -0700, Junio C Hamano wrote:
Show 8 quoted lines
> BTW, has anybody taken a look at this one? > > Subject: BUG: fetch incorrect interpretation of globing patterns in refspecs > Date: Thu, 24 Jul 2008 09:07:21 +0200 > Message-ID: <71295b5a0807240007k246973abj1897895d0d67bb6c@mail.gmail.com> > > If not, I think I probably need to take a look at this, reproducing and > possibly fixing, before applying non-fix patches.
I have been meaning to look at it for days, so I finally took a peek. I was able to reproduce the problem easily. I think it is (almost) as simple as the patch below. In the refspec parsing, we already require globs to come after '/', so this is the analagous check during match.
Unfortunately, this breaks t1020 (something about failing to clone HEAD it looks like, so probably it is some boundary case for matching just "*"). I don't have time to look further, and I will be out of touch until probably Sunday evening, so hopefully somebody else can run with it.
---
diff --git a/remote.c b/remote.c index 0d6020b..3ae0431 100644 --- a/remote.c +++ b/remote.c @@ -1108,7 +1108,8 @@ static struct ref *get_expanded_map(const struct ref *remote_refs, for (ref = remote_refs; ref; ref = ref->next) { if (strchr(ref->name, '^')) continue; /* a dereference item */ - if (!prefixcmp(ref->name, refspec->src)) { + if (!prefixcmp(ref->name, refspec->src) + && ref->name[remote_prefix_len] == '/') { const char *match; struct ref *cpy = copy_ref(ref); match = ref->name + remote_prefix_len;