Merge lp:~vila/udd/608563-requeue-all-of-type into lp:udd
Status: | Merged |
---|---|
Merged at revision: | 413 |
Proposed branch: | lp:~vila/udd/608563-requeue-all-of-type |
Merge into: | lp:udd |
Diff against target: |
203 lines (+121/-20) 2 files modified
icommon.py (+42/-20) tests.py (+79/-0) |
To merge this branch: | bzr merge lp:~vila/udd/608563-requeue-all-of-type |
Related bugs: |
Reviewer | Review Type | Date Requested | Status |
---|---|---|---|
James Westby | Approve | ||
Review via email: mp+53527@code.launchpad.net |
Description of the change
Basically the fix is only:
@@ -686,7 +706,7 @@
- this_sig = self.failure_
+ this_sig = self.failure_
I've added some tangentially related tests and found a couple of minor issues while investigating which I'd like to land too.
Hmm, I was so surprised when I fixed the bug that I forgot to write a proper cover letter, here it is.
requeue_package.py --all-of-type <pkg_name> had a bug where all failed imports were requeued instead of only the ones with the same failure signature. This was due to a typo that makes the signature comparison always use the signature from <pkg_name>.
This patches fixes the typo and makes the signature more precise (since the full traceback is stored in the db this should only make things clearer and doesn't require updating the db itself).
There were also some dusty code in failure_signature that I got rid off (there can't be '\n' in the trace lines since they are produced by splitlines(), annotate and some more following across files revealed that this was needed when the traceback was acquired with readlines()).
Finally there was a bogus test against running_sentinel which *contains* a '\n' that should be ignored when comparing the signature from the db.