1709960 - [meta] Fix web platform tests for isolated process

Status: NEW

(GeckoView :: General, task, P2)

Description

[[:owlish] 🦉 PST]

Affected tests:

• web platform tests W(wpt)
• reftest W(Wr)

Once fixed, these should be enabled in automation (i.e. have the same tier in sandboxed builds as in non-sandboxed builds of the same type)

Comment 1

[Chris Peterson [:cpeterson]]

Moving isolated process bugs to the new GeckoView::Sandboxing component.

Comment 2

[Chris Peterson [:cpeterson]]

Tasks should have severity N/A.

Comment 3

[Olivia Hall [:olivia]]

Sent a try run to see the current state of these tests. I'll revisit and summarize later this week.

Comment 4

[Olivia Hall [:olivia]]

There seems to be a lot of intermittent failures in this WPT or an issue with how W was redefined post-SHIP and fission on the non-isolated process run, so I'm not sure if we have a good enough baseline.

However, W-nofis appears clear in the non-isolated process run, so I'll summarize off of those isolated process runs.

W-nofis

• Appear to be mostly around video or media failures, such as:
  • PROCESS-CRASH | application crashed [None] | /css/compositing/mix-blend-mode/mix-blend-mode-video-sibling.html
  • PROCESS-CRASH | application crashed [None] | /fetch/metadata/track.https.sub.html
  • PROCESS-CRASH | application crashed [None] | /navigation-timing/dom-interactive-media-document.html
  • PROCESS-CRASH | application crashed [None] | /html/semantics/the-button-element/command-and-commandfor/on-video-behavior.tentative.html
  • TEST-UNEXPECTED-NOTRUN | /content-security-policy/media-src/media-src-7_2.html | In-policy audio src - expected PASS
  • TEST-UNEXPECTED-TIMEOUT | /mediacapture-record/passthrough/MediaRecorder-passthrough.https.html | PeerConnection passthrough MediaRecorder receives VP8 after onstart with a video stream. - Test timed out
• (Not comprehensive, it is difficult to tell the intermittent failures from the true failures in this run.)

Comment 5

[Makoto Kato [:m_kato]]

All? WPT failures depend on remote media codec. We have to wait for fixing bug 1934009.

Comment 6

[Olivia Hall [:olivia]]

Once bug 1976487 syncs back into MC, we should have these issues catagorized under the search for bug: 1709960.

These were categorized under the W-nofis variant because the plain W variant had a lot of noise at the time of comparison.