Hi Abhradeep, On Wed, Aug 17, 2022 at 04:38:14PM -0400, Taylor Blau wrote: > > > Abhradeep -- let me know if this is something you want to look into. I > > > think it's a very worthwhile bug to fix, since it is definitely > > > trigger-able in the wild (notably, only with `git multi-pack-index write > > > --bitmap` without `--stdin-packs` and only under certain circumstances), > > > and not just limited to SHA-256 mode. > > > > > > If you are busy experimenting with CRoaring, that's no problem and I can > > > fix this up, too. Either way, it would be worth you and others weighing > > > in on which fix you think is worth pursuing. > > > > I will be happy to fix it but I can't work on it right now (neither on > > CRoaring) because I am currently preparing for my exam. I can continue > > my work after that (i.e. from 19 aug). If you feel it is getting too > > late then you can do this too. I am also thinking of writing a patch > > for bitmap specific test dump tool (as Johannes proposed previously). > > No problem. I wrote up some patches today myself that implement the > above fix. I haven't polished them up yet, but they are available here: > > https://github.com/ttaylorr/git/compare/master...ttaylorr:git:tb/bitmap-use-existing-preferred > > I want to add a more direct reproduction that works in both SHA-1 and > SHA-256 to demonstrate that these patches fix the issue. But in the > meantime, you can use Dscho's reproduction with these patches (based on > the tip of `master`) applied on top and observe that it passes > consistently. That is now done and I sent the resulting patch series to the list, which I'd encourage you to review here: https://lore.kernel.org/git/cover.1660944574.git.me@xxxxxxxxxxxx/T/#t Phew! Thanks, Taylor
