WebDriver/Meetings/2018-10-15

Agenda

ato: We had a meeting with the devtools last week.; Nothing practical came out of it, but good to clear the air and keep them appraised of what we intend to do.; They have the same impression of CDP that we do, that it would be good for Mozilla to implement it.
whimboo: Who was there?
AutomatedTester: [summarises who attended]; Can we have a meaningful discussion at TPAC about CDP? What can CDP do for us?; I will also have discussions with other vendors regarding this.
ato: But they have different reasons for wanting that does not necessarily overlap with what we want out of CDP.; jgraham ran some grep expressions on the Puppeteer source code and came up with some crude information about what parts of CDP is used: https://docs.google.com/document/d/1YHZ7FAhGhH_VUyxuj4auwOQu4EMKDBi3-RWPbjrjOxs/edit

whimboo: Should we prepare issues to raise at TPAC?
ato: Yes, good thing you mentioned.; We should add them to https://www.w3.org/wiki/WebDriver/2018-TPAC-F2F to ensure they are addressed.

ato: I’ve disabled ARMv7 HF builds to preparation of releasing from TaskCluster and making the geckodriver 1.0 release.; It’s not that we can’t support them on Travis, but we can’t yet on TaskCluster.; They are however quite easy to self-provide.

whimboo updated us on the intermittents of the last two weeks. Highlights are:

[fixed] Firefox startup hang for ASAN builds on Linux64 (Bug 1487243)
[fixed] Lots of shutdown hangs caused by VRListenerThreadHolder (Bug 1494490, Bug 1494556)
[needs fix by kris] (Perma ccov) Assertion failure: m_refCount == 1 or pages_.empty() (Bug 1485985)

AttributeError: 'Marionette' object has no attribute 'is_shutting_down' if Marionette kills application due to a timeout (1496897)

whimboo: We didn’t handle this intermittent in the most ideal way.; What can we do to improve this?
ato: I saw your needinfo but I simply didn’t have capacity.; Was also out two days last week.; But I believe AutomatedTester wrote a patch for this.
AutomatedTester: Yes.
whimboo: How can we improve this to ensure we look at it sooner?
AutomatedTester: It depends how often it happens.; I don’t think setting a needinfo is the first thing to do after there’s one or a cases.; I won’t drop everything to look at a very low intermittent.; But it was clouded by the fact that there were a lot of different intermittent bugs that it turned out were related.; It wasn’t until I duped the bugs that I saw a pattern and that this was more serious than it first appeared.
whimboo: Part of the problem is that we don’t monitor other test jobs, just Mn.; We have to trust the sheriffs to alert us when they think something might be a Marionette issue.

whimboo: We have to get crash handling for wdspec implemented (geckodriver support first)! Any update?
ato: I haven’t looked at it since we last spoke.; I hope that my window manipulation changes can land this week.; I will make this my next priority.

(Spoken status updates in bold.)