Details
-
Bug
-
Resolution: Fixed
-
P3: Somewhat important
-
None
-
6.3
-
None
Description
Here are some differences I notice in the qmake (coin) vs cmake test run procedure, regarding failed tests:
When a test fails then it re-runs up to 4 extra times, in order to decide if the failure was a flake (flaky test)
- 2 failures used to be enough to fail the integration. Now with cmake it seems that after 2 failures the test runs again and passes the 3rd time and the integration continues. For example search for tst_QSocks5SocketEngine::simpleConnectToIMAP in this qmake run and in this cmake run.
- We used to re-run only the test function that failed, for example tst_progname funcname; now cmake re-runs everything. It's not clear to me what should be done if on the 2nd run, a different function fails.
Attachments
Issue Links
- is blocked by
-
QTQAINFRA-4681 All tests are super flaky on Android
- Closed
- relates to
-
QTBUG-96844 Rationalize logging in QtTestLib
- Closed
-
COIN-728 Logging categories no longer enabled when re-running failed tests
- Closed
- resulted from
-
QTBUG-96345 tst_QSocks5SocketEngine::simpleConnectToIMAP() is flaky on Ubuntu
- Closed
- resulted in
-
QTQAINFRA-4687 Android test results are not written to influx
- Closed
- mentioned in
-
Page Loading...