- [gbrown] Starting to look at "new/modified test verification" - bug 1357513
- overall great idea, web-platform does this on upstream new tests, 10x per test
- keep it small, don't do dependent files, or run randomization/high resource programs in the background
- how to track the effectiveness of this:
- whiteboard tags for intermittents related to test changes
- build a dashboard for recent changes which are 5+ intermittents in a week
- [gbrown] Trying to enable eslint on more tests - bug 1357557
- will spend a few days hacking on this to see how easy this is to make a big sweep through the tree
- might be tough to whitelist existing failures or only run on new changes.
- [jmaher] Lifecycle of managing intermittents today
- 3 sources:
- overall OF (looking for missing stockwell tags)
- neglected
- needswork
- [gbrown] reduced OF to the last day or two to find most recent troublemakers
- is there value to retriggering to identify the root cause
- [jmaher] Roadmap for autoclassification in Q2/Q3 ? (how can we track what is auto classified vs not)
{
"from":"treeherder",
"select":[
{"aggregate":"count"},
{
"name":"is_auto",
"value":{
"when":{"in":[
{"literal":"autoclassified intermittent"},
"failure.notes.failure_classification"
]},
"then":1,
"else":0
},
"aggregate":"sum"
}
],
"edges":"bugs.bug_id",
"where":{"and":[
{"eq":{"failure.notes.failure_classification":"intermittent"}},
{"gt":{"last_modified":{"date":"today-week"}}},
{"ne":{"repo.branch.name":"try"}}
]},
"limit":10000
}