Resolve #3494332 "Try to pass"
Closes #3494332
Merge request reports
Activity
added 42 commits
-
cfc94414...e2b97e56 - 10 commits from branch
project:11.x - e2b97e56...05a3b239 - 22 earlier commits
- 362367f3 - Completely remove the timeout override.
- 429b455c - Retry interval to 1.
- 7c280c27 - Remove node session timeout override.
- 7e90a543 - Enable additional logging.
- a46314eb - Increase logging.
- 9aa6f62b - Bump Chrome version.
- 84fe5660 - Try enabling xvfb
- e9ade023 - Put resources back to a reasonable amount.
- 1dbf6ccf - Revert debug changes.
- 5ee34828 - Update assertion.
Toggle commit list-
cfc94414...e2b97e56 - 10 commits from branch
added 2 commits
94 94 SE_NODE_OVERRIDE_MAX_SESSIONS: "true" 95 95 SE_NODE_MAX_SESSIONS: "16" 96 96 SE_SESSION_RETRY_INTERVAL: "1" 97 SE_SESSION_REQUEST_TIMEOUT: "10" 98 SE_START_XVFB: "false" 97 SE_SESSION_REQUEST_TIMEOUT: "60" Self review. Whatever the timeout was set at before, it was having almost no effect because even a 300 second timeout would not result in a session getting picked up. And when it did time out, the test would just be skipped and we'd not know.
I increased it to 60 because just in case a test method does get stuck waiting we don't want tests to fail, but don't really want to set it higher because if we do get a similar or similar-ish problem with selenium again, I think we'd want to see the timeouts happen and not wait for minute. We can always tweak this later.
60 59 try { 61 60 return parent::initMink(); 62 61 } 63 catch (DriverException $e) { 64 if ($this->minkDefaultDriverClass === DrupalSelenium2Driver::class) { 65 $this->markTestSkipped("The test wasn't able to connect to your webdriver instance. For more information read core/tests/README.md.\n\nThe original message while starting Mink: {$e->getMessage()}"); 66 } 67 else { 68 throw $e; 69 } 70 } 71 62 catch (\Exception $e) { 72 $this->markTestSkipped('An unexpected error occurred while starting Mink: ' . $e->getMessage()); 63 $this->mink = NULL; -
can't RTBC, because this part is from me.
If it's now working reliably with the XVBF change then it might be OK to get this in. I think there is an option that we only commit the variable changes and do some more testing with this.
A comment on the mink reset might also be useful (to avoid breaking in tearDown() and hiding the real error)
What about contrib? A lot less jobs usually, but it looks like this starts to happen already with a pretty low amount of tests. We should probably create an issue to check the gitlab_templates variables?