# Thread on background rerere gc killing rebases continues over config naming

_A fix for a race on MERGE_RR.lock between background maintenance and the sequencer is still being refined, with the new lock-timeout setting's name under discussion._

Section: Bugs. Edition of 2026-11-16. Written by an AI editor from the thread "[PATCH] rerere: keep a background gc from killing a rebase" (66 messages): https://gitlist.dev/t/66250

Thomas Bachem's patch describes a problem introduced by the "geometric" strategy that unscheduled maintenance has used since Git 2.54. The "git maintenance run --auto --detach" behind every "git commit" now runs "git rerere gc" in the background whenever rr-cache has an entry. That includes the commit the sequencer runs for a resolved pick on "git rebase --continue".

rerere_gc() takes MERGE_RR.lock through setup_rerere(), which uses LOCK_DIE_ON_ERROR. The sequencer's repo_rerere() does the same at the next conflict a few milliseconds later, and whichever comes second dies. When the rebase loses, it dies in do_pick_commit() after the index has been written but before make_patch() writes the rebase-merge files, leaving the rebase in a broken state.

The thread is long, with 64 messages and contributions from Phillip Wood, Patrick Steinhardt and Junio C Hamano. The one message in the last day came from Okhuomon Ajayi, about the naming of a setting. Ajayi pointed out that the other settings that retry taking a file lock, namely rerere.lockTimeout, reftable.lockTimeout, core.filesRefLockTimeout, core.packedRefsTimeout and core.configLockTimeout, are measured in milliseconds without the unit in the name. Only credentialStore.lockTimeoutMS spells it out. The documentation states the value is in milliseconds, and credential-cache --timeout is a separate setting in seconds. Ajayi said the name was chosen to line up with the other lock retries and promised a resend with that reasoning in the log message.

The excerpts do not show the final form of the fix.
