Fixing flaky tests for windows 7 32bit#23480
Conversation
💚 Build Succeeded
Expand to view the summary
Build stats
Test stats 🧪
💚 Flaky test reportTests succeeded. Expand to view the summary
Test stats 🧪
|
|
@v1v , most of tests are failing consistently most likely due to are not considered. If we change the value |
windows stage is mandatory, but optional stages are not mandatory for PRs
|
Pinging @elastic/integrations (Team:Integrations) |
|
running the tests again /test |
|
/test |
| // under the License. | ||
|
|
||
| // +build !integration | ||
| // +build !integration, !386 |
There was a problem hiding this comment.
Are these tests failing in all 32-bit platforms or only in Windows?
There was a problem hiding this comment.
Could we explicitly skip the tests only for Windows 32 bits instead of for all 386 platforms? With a comment to know what they are skipped.
There was a problem hiding this comment.
sure, I have made the changes
|
|
||
| assert isinstance(udp["all"]["count"], int) | ||
|
|
||
| @unittest.skipIf(sys.platform == "win32", "Flaky test") |
There was a problem hiding this comment.
Flaky tests should have a open issue to address them in the future.
There was a problem hiding this comment.
I would have created one separately but I had around 40+ tests, it would have been hard to manage the tickets, I have taken screenshots with all the ones that kept on popping up
There was a problem hiding this comment.
Umm, does this test always fail in win32? If it always fail maybe instead of skipping it because of flakiness we can say in the comment that this test (or this feature) is not supported in win 32.
There was a problem hiding this comment.
the test does not always fail, it passed previously.
There was a problem hiding this comment.
If we plan to fix these flakiness at some moment we should have an issue for that, and reference it in the comment. Flaky tests can be hiding bugs.
But possibly it doesn't make much sense to dedicate a lot of effort to corner cases in 32-bits Windows 🙂
|
I've just added all the labels for all the /test |
|
/test |
|
|
||
| assert isinstance(udp["all"]["count"], int) | ||
|
|
||
| @unittest.skipIf(sys.platform == "win32", "Flaky test") |
There was a problem hiding this comment.
If we plan to fix these flakiness at some moment we should have an issue for that, and reference it in the comment. Flaky tests can be hiding bugs.
But possibly it doesn't make much sense to dedicate a lot of effort to corner cases in 32-bits Windows 🙂
|
/test |
1 similar comment
|
/test |
* enabling testing on windows 32 bit * update test file * fix * [CI] Optional stage instead the mandatory one windows stage is mandatory, but optional stages are not mandatory for PRs * fix int issue * ignore 386 architecture * skip tests * skip test * function redeclared * re write * add windows constraint * add tag * add comments Co-authored-by: Victor Martinez <VictorMartinezRubio@gmail.com> (cherry picked from commit 1b43637)
* enabling testing on windows 32 bit * update test file * fix * [CI] Optional stage instead the mandatory one windows stage is mandatory, but optional stages are not mandatory for PRs * fix int issue * ignore 386 architecture * skip tests * skip test * function redeclared * re write * add windows constraint * add tag * add comments Co-authored-by: Victor Martinez <VictorMartinezRubio@gmail.com> (cherry picked from commit 1b43637)
* enabling testing on windows 32 bit * update test file * fix * [CI] Optional stage instead the mandatory one windows stage is mandatory, but optional stages are not mandatory for PRs * fix int issue * ignore 386 architecture * skip tests * skip test * function redeclared * re write * add windows constraint * add tag * add comments Co-authored-by: Victor Martinez <VictorMartinezRubio@gmail.com> (cherry picked from commit 1b43637)
* enabling testing on windows 32 bit * update test file * fix * [CI] Optional stage instead the mandatory one windows stage is mandatory, but optional stages are not mandatory for PRs * fix int issue * ignore 386 architecture * skip tests * skip test * function redeclared * re write * add windows constraint * add tag * add comments Co-authored-by: Victor Martinez <VictorMartinezRubio@gmail.com> (cherry picked from commit 1b43637)
Will update soon
What does this PR do?
After investigation we see 2 main issues:
constant .... overflows intBecause the
intvalue is overflowed, the fix is to specify theint64type instead.expfmtpackage. This happens in function https://github.com/elastic/beats/blob/master/metricbeat/helper/prometheus/prometheus.go#L72When trying to read and parse the data, lines for example:
are not considered.
Why is it important?
Should fix/skip the following:

(etcd failing tests were previously fixed)
Python tests
Checklist
CHANGELOG.next.asciidocorCHANGELOG-developer.next.asciidoc.Related issues