From: Jeff King Date: Mon, 25 May 2026 10:02:15 GMT Subject: Re: [PATCH 1/3] commit-reach: deduplicate queue entries in paint_down_to_common Message-ID: <20260525100215.GB3868724@coredump.intra.peff.net> In-Reply-To: On Mon, May 25, 2026 at 09:53:09AM +0200, Kristofer Karlsson wrote: > Good catch Jeff! I think it's possible that I missed the flag cleanup case > here, but it's also possible that I got lucky and it worked anyway. Well, you did add it to ALL_FLAGS, so it might have been your subconscious making you lucky. :) > That said, I think the observation in the other email thread/commit is key > here. I will reply back in that one, but it seems like this can all be > simplified using Jeff's idea with an amortized O(1) solution by caching a > known non-stale entry in the queue, and thus becomes obsolete. I will post > a new patchset when the discussion slows down. Nifty, thanks. > As for general flag management, I will spend some more time thinking about it. > I don't fully trust static code analysis to work, but some cheap assertion > based model might give a nice trade-off. I think it would be really nice if we had per-operation flags kept outside of the structs completely. If you're a masochist, I fiddled around a bit with using a hash instead in this thread: https://lore.kernel.org/git/20250826055210.GA1031277@coredump.intra.peff.net/ It's sadly (but not surprisingly) quite slow. I do wonder how a slab would work there, but it would take a bit more surgery. We only allocate slab ids for commits, and we'd have to do so for all objects if we want to hold flags. Probably a dead-end, but it would be neat if all of these flag allocation worries just went away. -Peff