fail gracefully for pointless merge
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
jenkins-launchpad-plugin |
Confirmed
|
Medium
|
Unassigned |
Bug Description
This should not be happening:
DEBUG: Merging lp:~timo-jyrinki/ubuntu-ui-toolkit/qdocpaths to lp:ubuntu-ui-toolkit
Traceback (most recent call last):
File "/var/lib/
sys.
File "/var/lib/
return merge_and_
File "/var/lib/
target.
File "/usr/local/
branch.
File "<string>", line 4, in merge_from_
File "/usr/lib/
raise errors.
bzrlib.
Related branches
- PS Jenkins bot: Approve (continuous-integration)
-
Diff: 86 lines (+28/-15)2 files modifiedautoland.py (+7/-1)
tests/test_autoland.py (+21/-14)
Changed in jenkins-launchpad-plugin: | |
importance: | Undecided → Medium |
status: | New → Confirmed |
From https:/ /code.launchpad .net/~mrazik/ jenkins- launchpad- plugin/ stale-locks- take2/+ merge/166452 :
>> There is one issue, though -- fasttrack merges. I just noticed a pointless
>> merge and as far as I can see it is because the autolanding was triggered
>> while the other generic-land (which I started manually due to some unrelated
>> bug) was still running.
>> OTOH the problem is probably not introduced by this branch and explains why
>> I'm seeing these pointless merges every now and then.
>
> I was wrong about this. It is a newly introduced regression. OTOH solving this
> would most likely add yet another level of complexity and:
> 1. even if it happens nothing really bad happens (just the pointless merge
> error)
> 2. it shouldn't be happening very often as the generic-land is usually
> finished in 15 mins (the time between to triggering operations)
Right. The fasttrack MPs never call the autolanding job, they go straight to generic-land. As this new method looks for the MPs that were started from the autolanding job, it won't find a downstream build due to the upstream check.
I agree that we are safe for now as long as the second generic-land does not change the state of the MP from Merged to something else.