From: John Keeping <john@keeping.me.uk>
Date: Sun, 23 Jun 2013 14:55:03 GMT
Subject: Re: fast-import bug?
Message-ID: <20130623145503.GH4676@serenity.lan>
In-Reply-To: <m2ppvc29le.fsf@cube.gateway.2wire.net>

On Sun, Jun 23, 2013 at 07:19:25AM -0700, Dave Abrahams wrote:
> on Sun Jun 23 2013, John Keeping <john-AT-keeping.me.uk> wrote:
> > In this case, I think I do now understand why the mode is 0: in
> > parse_ls a new tree object is created and the SHA1 of the original is
> > copied in but the mode is left blank; clearly this should be set to
> > S_IFDIR when the SHA1 is non-null.
> >
> > I think the patch I now have is correct (and addresses the "copy from
> > root" scenario), but I need to spend some time understanding t9300 so
> > that I can add suitable test cases.
> 
> t9300?  

t/t9300-fast-import.sh in Git's source tree - it's where the tests for
fast-import live.

> Thanks; I'll try this one too.

Thanks.  I now have a patch series incorporating this which also adds a
few tests for handling of empty paths.  I'm sending it out in the next
few minutes.