Re: [PATCH 09/11] Provide API access to init_db()

> On Sun, 9 Mar 2008, Johannes Schindelin wrote:
> 
> > On Sat, 8 Mar 2008, Daniel Barkalow wrote:
> > 
> > > The caller first calls set_git_dir() to specify the GIT_DIR, and then 
> > > calls init_db() to initialize it. This also cleans up various parts of 
> > > the code to account for the fact that everything is done with GIT_DIR 
> > > set, so it's unnecessary to pass the specified directory around.
> > 
> > I think the most controversial part is that you change the way it is 
> > determined if the repository is a bare one.  It might have the same 
> > result (I do not have time to check it...), but the computation is 
> > performed differently.
> 
> It's actually doing the same computation; the difference is how the result 
> is reported. Mine commits to the repo being non-bare if it doesn't think 
> it's bare, which the current code leaves as "we don't know, assume 
> non-bare". But since we actually create a work tree if we guess it's 
> non-bare, it really is non-bare. So then the later code can assume it's 
> either 0 or 1, and not worry about -1 being equivalent to 0.