To help prioritizing writing tests, we could try hooking up the xenon logs used to generate flamegraphs (https://performance.wikimedia.org/xenon/logs/daily/), build a list of most called functions, and then figure out which of those are not covered.
Description
Description
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | Legoktm | T186300 Try to hook up xenon logs to coverage reports to get a list of most executed non-covered code | |||
| Resolved | Legoktm | T196031 Document CoverMe in relevant locations |
Event Timeline
Comment Actions
Finished: https://tools.wmflabs.org/coverme/
I think this is actually going to be pretty useful, e.g. https://tools.wmflabs.org/coverme/?repo=MediaWiki+core&type=load shows pretty obvious suggestions for improvement across ResourceLoader related code.
Because of the way the xenon logs work, it has a heavier bias towards entrypoint functions (e.g. MediaWiki::run()), but once those are covered they won't get in the way :)