Hi Junio, On Fri, 12 Aug 2022, Junio C Hamano wrote: > Victoria Dye <vdye@xxxxxxxxxx> writes: > > > Johannes Schindelin wrote: > > Just a heads-up: there was an issue with the process used to generate this > > version. ... > > Sorry to see that it caused two cycles of release process. Would it > have helped you if I had done things differently (other than "not > issuing maintenance releases that are not security relevant"), perhaps > giving a notice in advance by say a few days? Thank you for offering to accommodate Git for Windows' needs. In this instance, we tracked the problem down to https://github.com/gitgitgadget/git/issues/1321: When a rebase has a `pick` & `fixup` sequence and the `pick` failed and was skipped, then the `fixup` is still trying to amend `HEAD`, which is incorrect. In this instance, that wreaked quite the havoc, dropping most of the Windows-specific changes due to mistaking the initial fake merge [*1*] for a real merge. I offer a more verbose version of this analysis in https://github.com/git-for-windows/git/pull/3981#issuecomment-1212856115 Ciao, Dscho Footnote *1*: In Git for Windows, we use the "merging-rebase" strategy where each rebase starts with a fake merge of the previous commit history ("fake" because it uses `-s ours` to avoid taking any actual content changes). The idea is that the patches will be as clean and ready to submit to upstream as possible by rebasing them, yet still be able to fast-forward. For more details, see https://github.com/git-for-windows/build-extra/blob/ab941145aeef/shears.sh#L12-L18
