On Mon, Oct 28, 2024 at 04:35:22PM +0000, Caleb White wrote: > On Sun Oct 27, 2024 at 7:34 PM CDT, Taylor Blau wrote: > > Additionally, it appears that this patch breaks at least t0001-init.sh, > > but others as well. The relevant portion of -vx from t0001.46 is: > > > > --- 8< --- > > + git -C mainwt config worktree.useRelativePaths true > > + git -C mainwt worktree add --detach ../linkwt > > Preparing worktree (detached HEAD 89e2b84) > > HEAD is now at 89e2b84 gumby > > + git -C mainwt init --separate-git-dir ../seprepo > > Reinitialized existing Git repository in /home/ttaylorr/src/git/t/trash directory.t0001-init/seprepo/ > > + git -C mainwt rev-parse --git-common-dir > > fatal: repo version is 0, but v1-only extension found: > > relativeworktrees > > --- >8 --- > > > > Let's please fix this breakage, and then decide what we are going to do > > with the extension in the second patch relatively soon, as I worry that > > putting the original topic on master may have been a mistake if it is > > going to break worktree pruning on earlier Git versions. > > Thanks for catching this, the issue is that there's a bug in the > `initialize_repository_version()` function when reinitializing a > repository---the function doesn't check if there are other extensions > in use before downgrading the repository version to 0. > > This is a rare edge case, but I'll work on a fix. I am confused... are you saying t0001.46 is flaky? I don't think that it is, as it fails consistently for me with your patches applied when ran in a loop. I am definitely guilty of having sent broken patches to the list before that fail CI, usually when I have amended something I thought was trivial after already running the test suite on all patches, only to realize that it wasn't trivial after all and instead broken something. But please do be careful to thoroughly vet your patches before sending them to the list, as this sort of breakage can be disruptive. Thanks, Taylor
