Increase timeout when looking for user in list #31072
Conversation
Signed-off-by: Louis Chemineau <louis@chmn.me>
artonge
requested review from
a team,
nickvergessen,
juliushaertl and
come-nc
and removed request for
a team
|
Ironically the concerned test is failing in this PR... |
I think the problem is more that the selector sometimes does not match. Maybe it fails to switch to the disabled user tab or something? |
|
I know that on some test, not sure if it was this one, I once thought that the problem was the password timeout, because in some cases Nextcloud will open a pop-up to ask for your password to confirm some actions, but only if there have been some time since last time the password was entered. I was wondering if this was not the problem with the test, sometimes it would be too long after login and this pop-up would mess up with the selector. Just a theory though. (I thought about that because when attempting to do the same thing as the test by hand I had the pop-up in the place the test failed) |
|
Password timeout is 30 minutes |
I thought about that, but from looking at the element selection code, it looks like it would fail beforehand with an exception.
Could be, but as said nickvergessen the timeout is 30 minutes. Do you have the example of fixing a test by handling this popup? |
No, this was just an idea I had because once when trying to reproduce a failure by hand I had this password pop-up in the same place the test failed. But it may be a wrong idea if the timeout is 30minutes. |
|
Maybe start checking if other users are there, or check if the popup is shown, etc? |
|
FYI, this was an attempt, but it does not work |
My empirical findings show that https://github.com/nextcloud/server/blob/master/tests/acceptance/features/users.feature#L48 is causing half of the failures of acceptances tests in CI.
My assumption is that the server is not powerful enough and that a 100 seconds timeout is too short. But without more context like screenshots on failures, it is difficult to know for sure. Also, it is hard to reproduce locally as this does not fail all the time.
This PR simply increase the timeout to 150 seconds, hoping that this will fix the regular CI failures and help us having more trust in our CI tests.