Reinstate disabled MetricsPingScheduler tests
Categories
(Data Platform and Tools :: Glean: SDK, enhancement, P1)
Tracking
(Not tracked)
People
(Reporter: mdroettboom, Assigned: janerik)
Details
Attachments
(2 files)
As part of the work to fix intermittent glean tests failing, a handful of MetricsPingScheduler tests were disabled, since they were still failing.
We need to investigate why these tests were failing and restore them.
| Reporter | ||
Updated•7 years ago
|
Comment 1•7 years ago
|
||
| Reporter | ||
Updated•7 years ago
|
Updated•7 years ago
|
Comment 2•7 years ago
|
||
I am beginning to think now that the test isn’t intermittently failing (which I can control the frequency of fairly directly now through setting a delay in the background task that is calling stringMetric.set()). I think that the test is actually intermittently successfully exposing a bug because of how quickly the background task was writing. Being able to control how often it would show up though controlling how quickly the background task was writing got me to thinking that maybe there was a slight window in which we could write to a metric, and before the call to metricsPingScheduler.startupCheck() in Glean.initialize(). It looks like from when initialized is set to true in the function, to when we call .startupCheck() is several lines of code, where it looks like it may be possible to write to a metric before the window is closed and the ping built and sent.
Comment 3•7 years ago
|
||
Comment 4•7 years ago
|
||
This is suffering from some weird intermittent hangs on CI when testing that needs further investigation before I can merge and close this out.
Comment 5•7 years ago
|
||
Un-taking this and moving to M9
Updated•7 years ago
|
Updated•6 years ago
|
Updated•6 years ago
|
| Reporter | ||
Comment 6•5 years ago
|
||
Action: Determine if still relevant, and if not, remove code.
| Assignee | ||
Updated•5 years ago
|
Comment 7•5 years ago
|
||
| Assignee | ||
Updated•5 years ago
|
| Assignee | ||
Updated•5 years ago
|
Description
•