Merge lp:~xnox/ubiquity/fix-value-errors into lp:ubiquity
Status: | Merged |
---|---|
Approved by: | Colin Watson |
Approved revision: | 5654 |
Merged at revision: | 5681 |
Proposed branch: | lp:~xnox/ubiquity/fix-value-errors |
Merge into: | lp:ubiquity |
Diff against target: |
41 lines (+11/-5) 2 files modified
debian/changelog (+8/-0) ubiquity/plugins/ubi-partman.py (+3/-5) |
To merge this branch: | bzr merge lp:~xnox/ubiquity/fix-value-errors |
Related bugs: |
Reviewer | Review Type | Date Requested | Status |
---|---|---|---|
Colin Watson (community) | Approve | ||
Ubuntu Installer Team | Pending | ||
Review via email: mp+123727@code.launchpad.net |
Description of the change
It appears that previous fix to the ValueError was incomplete. Specifically the cleanup method is executed on abort and results in ValueError when run in ubiquity-dm session at the stepPartAsk page / "What would you like to do?".
So it seems like communication with debconf inside cleanup is also not-safe at the "pre-install" steps. But also I don't see a reason, why this should be communicated to debconf at the cleanup. Cleanup is a bit a catch all and is called at cancel, back-out to previous plugin and abort.
I am proposing to move grub device pre-seeding into the ok_handler which is executed only when confirming the installation / moving to the next step, while we still have debconf opened.
I still need to test manual partitioning & Qt frontend. But this seems like a better solution.
Colin, where there other reasons to run grub-installer/
This was moved to exit_ui_loops (and later to cleanup) due to bug 756920. I'm OK with you moving it back to ok_handler if you can demonstrate that that bug no longer occurs; it's certainly possible that it's since been fixed some other way, but it was a sufficiently bad bug that we should test it.