From: Eric Wong <normalperson@yhbt.net>
Date: Fri, 23 Oct 2009 06:38:15 GMT
Subject: Re: [bug][bisected] git-svn with root branches
Message-ID: <20091023063815.GA15445@dcvr.yhbt.net>
In-Reply-To: <c376da900910220824g2948dc2sa1156bda59b49405@mail.gmail.com>

Adam Brewster <adambrewster@gmail.com> wrote:
> > I'm not sure why Adam decided the leading slash needed to be removed for
> > the git refspec.  That said, the globbing/branching code still makes me
> > want to hide under a rock and I'm generally afraid to touch it.
> > I'll wait for Adam to chime in since he's braver than I am :)
> >
> How's this for brave:  I'm not sure why I did that either.
> 
> Looking at it again it seems correct for the leading / to be ignored because
> it has no meaning.  What's the difference between the above config and
> "branches = *:refs/remotes/*" (besides the fact that one works and one
> doesn't)?  In both cases I think I've asked for all of the top-level
> directories to be branches.  But that has nothing to do with the rest of the
> patch, so it shouldn't have been included.

<snip>

> That is, continue disregard the / on the actual input because it means
> nothing, and add a / in places where the code will crash if it's not there.
> Even better would be to find out _why_ $path and $local_ref need a leading /
> and fix it that way, but that's more work that I don't have time for right
> now.

Actually, SVN 1.5+ does not like the leading "/" and trips on an
assertion failure on all cases except a standalone "/".  That is, "/*"
continues to work since we send "/" to SVN::Ra::get_log().

-- 
Eric Wong