From: Junio C Hamano <gitster@pobox.com>
Date: Fri, 07 Jun 2013 19:46:25 GMT
Subject: Re: What's cooking in git.git (Jun 2013, #03; Thu, 6)
Message-ID: <7vwqq5snzi.fsf@alter.siamese.dyndns.org>
In-Reply-To: <20130607191643.GA31625@goldbirke>

SZEDER Gábor <szeder@ira.uka.de> writes:

>> Now I do not recall suggesting it, and you (and I today after 2
>> years) may disagree with the rationale, but at least we can read
>> what was the "intended" meaning, I think.
>
> See
>
>   http://thread.gmane.org/gmane.comp.version-control.git/167728/focus=168838
>
> I still agree with the rationale,...

Thanks for a pointer.  I think what I was suggesting was slightly
different in that I was hoping to see a single helper that knows to
complete to object names (possibly including trees/blobs with the
treeish:path notation), ranges, and pathnames (not treeish:path
notation) until it sees a "--" and then complete only to pathnames.

It can be improved by teaching the unified one that some command
like "log" can never take treeish:path but only committishes, some
command take individual object names but never ranges, and/or
details like that.  But I still agree that "git log HEAD:dir<TAB>"
that completes to a blob or a tree object name is not an issue
(because what was before <TAB> cannot possibly name anything "git
log" would appreciate), even though it might not be ideal.