Re: [PATCH 0/16] make prune mtime-checking more careful

From: Jeff King <peff@peff.net>
Date: Oct 6, 2014, 01:42 UTC
Message-ID: <20141006014249.GA11495@peff.net>
In-Reply-To: <xmqqbnprjx0t.fsf@gitster.dls.corp.google.com>

On Sat, Oct 04, 2014 at 03:22:10PM -0700, Junio C Hamano wrote:

> This applied on top of 'maint' (which does have c40fdd01) makes the
> test #9 (prune: do not prune detached HEAD with no reflog) fail.

I'll fix the bone-headed error returns that René noticed and double check that they were the complete culprit in the test failure you saw (and not just masking some other problem).

> If we merge 'dt/cache-tree-repair' (which in turn needs
> 'jc/reopen-lock-file') to 'maint' and then apply these on top, the
> said test passes.  But I do not see an apparent reason why X-<.

I suspect it's an unintended interaction that just happens to let my bogus code code the right thing, but I'll double check on that, too.

Thanks both of you for spotting the problems.

-Peff