Re: [PATCH 6/7] Bisect: factorise "bisect_{bad,good,dunno}" into "bisect_state".

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Oct 16, 2007, 11:20 UTC
Message-ID: <Pine.LNX.4.64.0710161218260.25221@racer.site>
In-Reply-To: <200710160828.05137.chriscool@tuxfamily.org>

Hi,

[finally, a technical argument on "dunno". Maybe we really need git-bikeshed@vger.kernel.org? ;-)]

On Tue, 16 Oct 2007, Christian Couder wrote:

Show 15 quoted lines

> Le lundi 15 octobre 2007, Johannes Schindelin a ?crit :
> >
> > On Mon, 15 Oct 2007, Christian Couder wrote:
> > >
> > > But the new "bisect_state" takes one more argument, because the first
> > > one must be "good" "bad" or "dunno".
> > >
> > > So when there is only one argument HEAD is used, and when there are 
> > > 2 arguments, $2 is used as the good|bad|dunno rev.
> >
> > Ah, that explains it!  But do you not need to do 
> > "2,bad|2,good|2,dunno" in that case?  Or even better: "2,*"?
> 
> Perhaps it would be an improvement at least in speed for "2,good" or 
> "2,dunno".

I see: the later case catches them.  Colour me satisfied with your patch.

> I wanted to keep exactly the same processing as there was before, in case 
> something like "git bisect good v1.5.3.3..v1.5.3.4" was supported. But it 
> seems it doesn't work. I don't know if it's a bug or a feature.

I think v1.5.3.3..v1.5.3.4 expands to ^v1.5.3.3 v1.5.3.4, which might explain what you are experiencing.

OTOH "git bisect good v1.5.3.3..v1.5.3.4" does not make sense. bisect assumes that all ancestors of a good commit are good, too.

Ciao, Dscho