Re: Bug: git branch -D can be used to delete branch which is currently checked out - Part 2

From: Marcus Kida <marcus.kida@gmail.com>
Date: Mar 10, 2016, 22:43 UTC
Message-ID: <02FB9587-5849-4B3A-A199-6483377C3EED@gmail.com>
In-Reply-To: <20160310224158.GA2772@sigill.intra.peff.net>

Fair enough,

thank you. I’m going to take a look at the previous threads. I’d also be keen to help working on those issues.

Show 27 quoted lines

> On 11 Mar 2016, at 9:41 AM, Jeff King <peff@peff.net> wrote:
> 
> On Fri, Mar 11, 2016 at 05:30:00AM +1100, Marcus Kida wrote:
> 
>> thank you for the feedback.
>> I will fix this, test it and send a patch.
> 
> Unfortunately, I think this issue is a little more complicated.
> 
> There's some prior discussion in
> 
>  http://thread.gmane.org/gmane.comp.version-control.git/284022
> 
> and
> 
>  http://thread.gmane.org/gmane.comp.version-control.git/276456/focus=276506
> 
> The latter, in particular, shows a case where this approach will do the
> wrong thing. The fundamental issue is that refs are potentially stored
> in _two_ places: the filesystem, and the packed-refs file. And the
> latter is always case-sensitive, while the former sometimes is and
> sometimes isn't. But because the storage all happens behind the scenes,
> the user has no way of reliably disambiguating (e.g., does "foo" refer
> to your checked-out "FOO", or are you intentionally trying to delete an
> extraneous "FOO" that ended up in the packed-refs file?).
> 
> -Peff