2026-02-06 11:12 UTC
Replies (3)
-
@Natanox@chaos.social 2026-02-06 12:06
@zekjur @RandamuMaki I do wonder though, does gitlab or forgejo behave differently?
-
@varx@cybersecurity.theater 2026-02-06 20:59
@RandamuMaki @zekjur Amusingly, you seem to have conflated git and GitHub here. :-) (It's a git issue, not a GH issue.)
-
@taylor@social.axfive.net 2026-02-07 04:27
@RandamuMaki @zekjur This was not caused by GitHub. The GitHub PR merged without issue (because it's not patch-based), it was when the Debian maintainer exported it as a patch and applied it that the problem appeared. patch is the problem, and git-am inherits the problem from patch. edit: Just to verify (I definitely don't want to unfairly not blame Microsoft; I love blaming Microsoft), I tested it out on the command line with just git tools, and got the same result. If GitHub is generating bad patches that need escaping, so does git format-patch: > git checkout e285b3c47b2fc7a4cdaba9a427d2958410e64620 HEAD is now at e285b3c4 Fix i3bar workspace buttons for primary screen (#6564) > grep sleep i3bar/src/workspaces.c > git format-patch e285b3c47b2fc7a4cdaba9a427d2958410e64620~..e285b3c47b2fc7a4cdaba9a427d2958410e64620 0001-Fix-i3bar-workspace-buttons-for-primary-screen-6564.patch > git checkout e285b3c47b2fc7a4cdaba9a427d2958410e64620~ Previous HEAD position was e285b3c4 Fix i3bar workspace buttons for primary screen (#6564) HEAD is now at 3a6f17d7 t/289-ipc-shutdown-event: remove AnyEvent timers > git am 0001-Fix-i3bar-workspace-buttons-for-primary-screen-6564.patch Applying: Fix i3bar workspace buttons for primary screen (#6564) > grep sleep i3bar/src/workspaces.c sleep(1); > This is an inherent problem with in-band signaling. The patch format is simply poorly structured for anything containing non-patch data, unless the non-patch data is specially and deterministically formatted to avoid collision. As-is, git-format-patch and GitHub's patch output can't do anything about it without changing the commit message (which is probably a better solution, given that the current patch breaks both the patch application and the commit message).