During the wb_terms leftover hike (https://phabricator.wikimedia.org/project/view/5012/) we think we noticed an increase in random browser tests failing.
It is unclear whether or not this is due to the fact that we were working with an unusually large amount of patches at the same time, thus putting more pressure on the CI servers. Or if this is due to something else.
- Are there any ways to improve this?
- Are there any ways for a failed browser test to be retried?
Example output:
14:11:38 [chrome 73.0.3683.75 Linux #0-5] 10 passing (1m 47.5s) 14:11:38 [chrome 73.0.3683.75 Linux #0-5] 1 failing 14:11:38 [chrome 73.0.3683.75 Linux #0-5] 14:11:38 [chrome 73.0.3683.75 Linux #0-5] 1) init indicates loading while app gathers its data 14:11:38 [chrome 73.0.3683.75 Linux #0-5] waitUntil condition timed out after 10000ms 14:11:38 [chrome 73.0.3683.75 Linux #0-5] Error: waitUntil condition timed out after 10000ms 14:11:38 [chrome 73.0.3683.75 Linux #0-5] at Context.it (/workspace/src/extensions/Wikibase/client/data-bridge/tests/selenium/specs/init.js:68:11) 14:11:38 ------------------------------------------------------------------ 14:11:38 [chrome 73.0.3683.75 Linux #0-6] Spec: /workspace/src/extensions/Wikibase/client/data-bridge/tests/selenium/specs/permissionChecks.js 14:11:38 [chrome 73.0.3683.75 Linux #0-6] Running: chrome (v73.0.3683.75) on Linux
Example patches:
https://gerrit.wikimedia.org/r/c/mediawiki/extensions/Wikibase/+/632740
https://gerrit.wikimedia.org/r/c/mediawiki/extensions/Wikibase/+/632899
https://gerrit.wikimedia.org/r/c/mediawiki/extensions/Wikibase/+/632695
https://gerrit.wikimedia.org/r/c/mediawiki/extensions/Wikibase/+/632727
https://gerrit.wikimedia.org/r/c/mediawiki/extensions/Wikibase/+/632649