From: Felipe Contreras Date: Sun, 08 Apr 2012 13:47:04 GMT Subject: Re: [PATCH 02/12] completion: simplify __gitcomp Message-ID: In-Reply-To: <20120408132704.GB4103@burratino> On Sun, Apr 8, 2012 at 4:27 PM, Jonathan Nieder wrote: > Felipe Contreras wrote: > >> So we should never make any cleanups because some custom completion >> functions might theoretically break? > > No, tasteful cleanups that don't break custom completion functions are > very welcome. > > Also if you have evidence that this is definitely (not just possibly) > only a theoretical problem, then it would not be a regression.  But I > actually find it likely that custom completion code would have > imitated the old completion code in git that used the suffix argument, > so... Show me an example. If you do something like: __gitcomp "foo bar opt=" "" "" "suffix" The result would be 'foo suffix' 'bar suffix' 'opt=suffix'. How could that *possibly* be useful? > Can you remind me what the benefit that the user is getting from this > patch Absolutely none. > in exchange for us breaking their tab completion? Whoa! Breaking their tab completion? Where? Can you show me some evidence? >  The title > ("simplify __gitcomp") tells me exactly nothing about its impact > except that it is perhaps supposed to result in no behavior change. That is correct. No functional changes. I would need to update the tests if something changed, right? Cheers. -- Felipe Contreras