fix(forge): the chain tiers accept whatever the canonical Forgejo holds #776
Loading…
Reference in a new issue
No description provided.
Delete branch "fix/forge-receive-fsck"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
What
Both tier configs gain
[git.config] receive.fsckObjects = false/transfer.fsckObjects = false.code.sean.farmtookillumos-omniosin over a fetch (no fsck); a push mirror arrives throughreceive-pack(fsck on), and one imported commit with a malformed author email hadcode.packet.townreject the whole 618 MB repository. The tiers are backups of the canonical instance, not gatekeepers of its history.Both hosts re-converged with the change; the
illumos-omnioshop-1 sync then lands and hop 2 follows on commit.https://claude.ai/code/session_01XdBRc9CHgFhAxCyHbdVtt6
The first dry run of the Codeberg mirror teardown found illumos-omnios with no copy on the chain: code.packet.town rejected its hop-1 push with remote: error: object 447603b5...: badEmail: invalid author/committer line remote: fatal: fsck error in packed object code.sean.farm took that history in over a fetch, which git does not fsck. A push mirror arrives through receive-pack, which does, and one imported commit with a malformed author email had the receiving tier reject the whole 618 MB repository. Both tiers are backups of the canonical instance, not gatekeepers of its history, so their app.ini now carries [git.config] receive.fsckObjects = false transfer.fsckObjects = false in both converge scripts. Re-converged both hosts; the hop-1 sync for illumos-omnios then lands, and hop 2 follows on commit. Claude-Session: https://claude.ai/code/session_01XdBRc9CHgFhAxCyHbdVtt6c337064ccb1a7164893c