47 bytes, text/x-phabricator-request
|Details | Review|
for the sake of being able to change it upstream without changes in m-c in https://github.com/web-platform-tests/wpt/issues/7598 and if we ever actually implement https://github.com/web-platform-tests/wpt/issues/7135
Assignee: nobody → geoffers+mozilla
Status: NEW → ASSIGNED
Depends on D13282
https://treeherder.mozilla.org/#/jobs?repo=try&revision=9f1462bf64454bb510c98b4f5642f50369772471 ran this through try (reftest wpt only, given that's all this should affect), and Win7 seems to fail when trying to encode images when converting the canvas to a data URL. Myself and James' assumption here is that it's hitting OOM, though this is a bit surprising given all it should really be changing is adding one extra layer of indirection (an extra Map) for the screenshots. Anyone got any advice on how to start figuring this out?
We're still failing on Win7, despite now listening for memory-pressure (and not getting any notifications). See https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=216615156&repo=try&lineNumber=8639: 16:38:13 INFO - PID 5260 | [Parent 4812, Main Thread] WARNING: NS_ENSURE_SUCCESS(rv, rv) failed with result 0x80004005: file z:/build/build/src/dom/base/ImageEncoder.cpp, line 407 16:38:13 INFO - PID 5260 | [Parent 4812, Main Thread] WARNING: NS_ENSURE_SUCCESS(rv, rv) failed with result 0x80004005: file z:/build/build/src/dom/html/HTMLCanvasElement.cpp, line 758 16:38:13 INFO - TEST-UNEXPECTED-ERROR | /css/CSS2/normal-flow/max-width-106.xht | [Exception... "Failure" nsresult: "0x80004005 (NS_ERROR_FAILURE)" location: "JS frame :: chrome://marionette/content/reftest.js :: toBase64 :: line 262" data: no] Note these failures are Win7 only. :njn or :gsvelto, :jgraham suggested you might be able to help with figuring out the cause of these failures?
To be clear, what I meant is that they might know if we expect to get memory-pressure notifications here if we happen to be going above the per-process limit on win32, and whether the way the observer is wired up is correct. It is still possible that the hypothesis that this error is OOM is entirely false.
The machinery that detects low-memory scenarios polls at a very slow rate (once per second) and only when the user is interacting with Firefox so it's unlikely that you'll get a memory-pressure event in automation. Looking at the log I see the error in ImageEncoder::ExtractDataInternal() but since there's not much info apart from that I'd sprinkle that call with printouts to figure out which one of the called functions is failing and then dig down to the source.
I don't have anything to add beyond what gsvelto said.
Pushed by email@example.com: https://hg.mozilla.org/integration/autoland/rev/daf05866f9dd Make width/height of reftests configurable over Marionette r=jgraham,ato
Created web-platform-tests PR https://github.com/web-platform-tests/wpt/pull/14821 for changes under testing/web-platform/tests
Pushed by firstname.lastname@example.org: https://hg.mozilla.org/integration/autoland/rev/0a9c477577de Make width/height of reftests configurable over Marionette r=jgraham,ato
You need to log in before you can comment on or make changes to this bug.