Race condition on builders result in 100% fail rate
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
ubuntu-drivers-common (Ubuntu) |
Invalid
|
Undecided
|
Unassigned | ||
Jammy |
Fix Released
|
Undecided
|
Unassigned |
Bug Description
[ Impact ]
ADT testing fails on arm builders due to a racy test
[Fix]
Split the test avoiding the race.
[ Test Plan ]
Run ADT tests in PPA or locally with:
PYTHONPATH=. python3 tests/run test_ubuntu_
The error message to look for is:
=======
FAIL: test_system_
system_
-------
Traceback (most recent call last):
File "/usr/lib/
return func(*newargs, **newkeywargs)
File "/home/
self.
AssertionError: True is not false
[ Where problems could occur ]
during ADT testing on builders
[ Other Info ]
Cause unknown. It sometimes fails when locally on x86 machine but only in Jammy container. Jammy package ran in Noble container does not fail.
summary: |
- Race condition on arm builders result in 100% fail rate + Race condition on builders result in 100% fail rate |
description: | updated |
description: | updated |
Changed in ubuntu-drivers-common (Ubuntu): | |
status: | New → Invalid |
description: | updated |
> + # Race condition may happen and the test fails. This only happens in Jammy containers.
> + # Disable the test for now.
> + return
I don't think it's appropriate to disable a test without an analysis that considers what it was testing, how to mitigate the gap created by disabling the test or an explanation of why it isn't necessary to run the test.
We also need to make arrangements to ensure that a future SRU or security update of this package will make the same mitigation.
In the lack of such an analysis or a proper fix, if it only sometimes fails, wouldn't it be better to just retry a few times to land the SRU?