Re: [PATCH 5/7] Makefile: add 'check-sort' target

From: Jeff King <peff@peff.net>
Date: Mar 17, 2021, 22:01 UTC
Message-ID: <YFJ8KpseGSJdd3Am@coredump.intra.peff.net>
In-Reply-To: <CAPig+cRk-fCLFtug47w3trYQFiPzZ3bD4g11Qu4fTd3Fh6_bCQ@mail.gmail.com>

On Wed, Mar 17, 2021 at 05:48:18PM -0400, Eric Sunshine wrote:

Show 13 quoted lines

> On Wed, Mar 17, 2021 at 1:34 PM Jeff King <peff@peff.net> wrote:
> > TBH, I'm a little on the fence on whether automatically checking this is
> > even worth the hassle. Doing the make function above was a fun
> > diversion, but already I think this discussion has taken more time than
> > people actually spend resolving conflicts on unsorted Makefile lists.
> 
> I had the same reaction. Like you, I jumped in for the fun diversion.
> It allowed me to flex my Perl muscle a bit which has atrophied, but an
> out-of-order item here and there is such a minor concern, especially
> since they don't impact correctness, that I worry that such a CI job
> would be more hassle than it's worth. Making the feedback loop
> tighter, as discussed elsewhere in this thread, makes the idea of the
> automated check a bit more palatable.

There's an implication in what both of us said that I think is worth calling out explicitly: I would not feel the same about a problem that impacts the correctness of the resulting code. E.g., if the list of builtins were used with a binary search, then an unsorted list would produce the wrong result. And that would be worth testing.

It just seems that the stakes here are much lower.

-Peff