From: Phillip Wood Date: Tue, 13 Oct 2026 15:49:39 GMT Subject: Re: [PATCH v3 2/2] merge: remember conflict labels Message-ID: <0e8e5042-9b0c-447f-a320-a3d58d364282@gmail.com> In-Reply-To: On 12/10/2026 16:09, Junio C Hamano wrote: > Phillip Wood writes: >> On 10/10/2026 01:24, Junio C Hamano wrote: >>> >>> This function is called once for each conflicted path, which means >>> we would read the "merge.renormalize" configuration variable and the >>> MERGE_LABELS file, both of which will stay constant during a single >>> conflicted "checkout -m". The issue is shared with the original, >>> but looking up the same configuration variable repeatedly would be >>> helped with in-core configset cache. Compared to that, the overhead >>> added by this patch is to open the same unchanging file, read & parse, >>> allocate and deallocate. >>> >>> Perhaps we want to have another preliminary [PATCH 1.5/2] before >>> this step to allow setting these "per invocation constants" once to >>> be reused? Then step [PATCH 2/2] can read labels in the prepare >>> phace just once, use it from the data structure in >>> checkout_merged(), and free them in release phase when we are done. >> >> That's a good idea, I'll re-roll using the diff below (I assume you're >> happy for me to add your sign off?) > > You assumed correctly, but if you ask me, I would say that it > probably does not count much as coauthorship than a mere helped-by. As the new patch is a very lightly edited version of your diff I've made you the author. Thanks Phillip