Re: [PATCH 0/5] refs: remove functions without ref store

From: Junio C Hamano <gitster@pobox.com>
Date: May 7, 2024, 15:46 UTC
Message-ID: <xmqqwmo5sjyq.fsf@gitster.g>
In-Reply-To: <ZjnKfllMW9rGgWPL@tanuki>

Patrick Steinhardt <ps@pks.im> writes:

Show 8 quoted lines

>> What I had in mind was a lot more stupid like the attached.  For
>> illustration purposes, I just did only one, but you got the idea.
>> Thanks to "#if 0", the compilation will fail, the compiler telling
>> the developer "resolve_ref_unsafe()? what are you talking about?",
>> and the developer will grep for that name to find the hint at the
>> end.
>
> That works for me, too. Thanks!

And if I replace 0 with USE_IMPLICIT_MAIN_REFSTORE (which is never defined anywhere, and only serves for documentation purposes with its rather explicit name), you get an exact implementation of what I meant by "an approach similar to USE_THE_INDEX_MACROS".

Having said all that, as I expect that the topic itself will go through smoothly down to 'master' (once a topic hits 'next', unless it still has dubious parts, it usually takes 7 calendar days before it goes to 'master') because the other parts of the series are fairly straight-forward, we do not need anything more than just removing them, without the helpful documentation in this case, especially given that the adjustment other people need to do is very mechanical (and the recipe is already in the coccinelle step). They can notice that their new call to resolve_ref_unsafe() does not compile, grep for resolve_ref_unsafe and find nothing, and then they'll know to use "log -Sresolve_ref_unsafe" to find what series removed it, what happened to its old callers, and then adjust their calls.

Thanks.