Revision not present errors due to push creating stacked 2a repositories with format 6 (unstacked) branches
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Bazaar |
Confirmed
|
High
|
Unassigned |
Bug Description
Branches are sometimes created with format 6 even though the initial push code thought it would be stacked and seeded it like a stacked branch.
Workarounds
===========
Before pushing make sure branches are in branch 7 format (run 'bzr upgrade --2a' on them)
After pushing, if a branch is affected
bzr upgrade --2a sftp://
python
import bzrlib.branch
b = bzrlib.
b.set_stacked_
Causes
======
Still being analyzed
Details
======
Somehow several different users with format 2a repositories are pushing broken branches to Launchpad. Specifically they have created stacked repositories, but format 6 branches, so when accessing history outside of the stacked repository the user (and Launchpad internals like the scanner) get RevisionNotPresent errors. This may be caused by #390431.
Example branches affected by this:
lp:~danilo/launchpad/bug-389391
lp:~julian-edwards/launchpad/team-join-sub-email-bug-382687
lp:~jml/launchpad/fix-run-all
summary: |
- Revision not present errors in Launchpad's scanner + Revision not present errors due to push creating stacked 2a repositories + with format 6 (unstacked) branches |
description: | updated |
Changed in bzr: | |
status: | Triaged → Confirmed |
tags: | added: check-for-breezy |
tags: | removed: check-for-breezy |
Changed in brz: | |
status: | New → Triaged |
importance: | Undecided → High |
no longer affects: | brz |
The traceback shows it is failing in iter_reverse_ revision_ history, on a revision from PQM. I assume these are stacked branches. Perhaps the stacked-on branch isn't branch/repository isn't working properly?
Hmm:
$ bzr revision-history lp:~jml/launchpad/fix-run-all pository( \'lp-1404727608 15824:/ //~jml/ launchpad/ fix-run- all/.bzr/ repository\ ')".')
bzr: ERROR: Server sent an unexpected error: ('error', 'Revision {<email address hidden>} not present in "CHKInventoryRe
Same via SFTP.
Oh, interesting... "bzr info -v" on that branch shows:
branch: Branch format 6
So, I think this is a recurrence of a bug we used to have where a stacked repository gets pushed up, then a format 6 (and thus unstacked) branch gets created on that repository.