Closed Bug 1275094 Opened 3 years ago Closed 3 years ago

opaque-fractional-displayport reftests are going to permafail on Android when Gecko 49 merges to Aurora

Categories

(Core :: Graphics: Text, defect, blocker)

Unspecified
Android
defect
Not set
blocker

Tracking

()

RESOLVED FIXED
mozilla50
Tracking Status
firefox47 --- unaffected
firefox48 --- unaffected
firefox49 --- fixed
firefox50 --- fixed

People

(Reporter: RyanVM, Assigned: snorp)

References

Details

(Whiteboard: [gfx-noted])

Only on Android.

https://treeherder.mozilla.org/logviewer.html#?job_id=20974666&repo=try#L5067

12:49:55     INFO -  REFTEST TEST-START | http://10.0.2.2:8854/tests/layout/reftests/async-scrolling/opaque-fractional-displayport-1.html
12:49:55     INFO -  REFTEST TEST-LOAD | http://10.0.2.2:8854/tests/layout/reftests/async-scrolling/opaque-fractional-displayport-1.html | 751 / 805 (93%)
12:49:55     INFO -  REFTEST TEST-LOAD | about:blank | 751 / 805 (93%)
12:49:55     INFO -  REFTEST TEST-UNEXPECTED-FAIL | http://10.0.2.2:8854/tests/layout/reftests/async-scrolling/opaque-fractional-displayport-1.html | failed reftest-opaque-layer: <div class="transparent-overlap-of-fractional-edge reftest-opaque-layer"> is not part of an opaque layer
12:49:55     INFO -  REFTEST TEST-END | http://10.0.2.2:8854/tests/layout/reftests/async-scrolling/opaque-fractional-displayport-1.html
12:49:55     INFO -  REFTEST INFO | Saved log: START http://10.0.2.2:8854/tests/layout/reftests/async-scrolling/opaque-fractional-displayport-1.html
12:49:55     INFO -  REFTEST INFO | Saved log: [CONTENT] Setting displayport to <x=0, y=0, w=200, h=300.2>
12:49:55     INFO -  REFTEST INFO | Saved log: [CONTENT] OnDocumentLoad triggering AfterOnLoadScripts
12:49:55     INFO -  REFTEST INFO | Saved log: Initializing canvas snapshot
12:49:55     INFO -  REFTEST INFO | Saved log: DoDrawWindow 0,0,800,1000
12:49:55     INFO -  REFTEST INFO | Saved log: [CONTENT] RecordResult fired
12:49:55     INFO -  REFTEST INFO | Saved log: RecordResult fired
12:49:55     INFO -  REFTEST INFO | Saved log: START about:blank
12:49:55     INFO -  REFTEST INFO | Saved log: [CONTENT] OnDocumentLoad triggering AfterOnLoadScripts
12:49:55     INFO -  REFTEST INFO | Saved log: Initializing canvas snapshot
12:49:55     INFO -  REFTEST INFO | Saved log: DoDrawWindow 0,0,800,1000
12:49:55     INFO -  REFTEST INFO | Saved log: [CONTENT] RecordResult fired
12:49:55     INFO -  REFTEST INFO | Saved log: RecordResult fired
12:49:55     INFO -  REFTEST TEST-START | http://10.0.2.2:8854/tests/layout/reftests/async-scrolling/opaque-fractional-displayport-2.html
12:49:55     INFO -  REFTEST TEST-LOAD | http://10.0.2.2:8854/tests/layout/reftests/async-scrolling/opaque-fractional-displayport-2.html | 752 / 805 (93%)
12:49:55     INFO -  REFTEST TEST-LOAD | about:blank | 752 / 805 (93%)
12:49:55     INFO -  REFTEST TEST-UNEXPECTED-FAIL | http://10.0.2.2:8854/tests/layout/reftests/async-scrolling/opaque-fractional-displayport-2.html | failed reftest-opaque-layer: <div class="transparent-overlap-of-fractional-edge reftest-opaque-layer"> is not part of an opaque layer
12:49:55     INFO -  REFTEST TEST-END | http://10.0.2.2:8854/tests/layout/reftests/async-scrolling/opaque-fractional-displayport-2.html
12:49:55     INFO -  REFTEST INFO | Saved log: START http://10.0.2.2:8854/tests/layout/reftests/async-scrolling/opaque-fractional-displayport-2.html
12:49:55     INFO -  REFTEST INFO | Saved log: [CONTENT] Setting displayport to <x=0, y=-100.2, w=200, h=300.2>
12:49:55     INFO -  REFTEST INFO | Saved log: [CONTENT] OnDocumentLoad triggering AfterOnLoadScripts
12:49:55     INFO -  REFTEST INFO | Saved log: Initializing canvas snapshot
12:49:55     INFO -  REFTEST INFO | Saved log: DoDrawWindow 0,0,800,1000
12:49:55     INFO -  REFTEST INFO | Saved log: [CONTENT] RecordResult fired
12:49:55     INFO -  REFTEST INFO | Saved log: RecordResult fired
12:49:55     INFO -  REFTEST INFO | Saved log: START about:blank
12:49:55     INFO -  REFTEST INFO | Saved log: [CONTENT] OnDocumentLoad triggering AfterOnLoadScripts
12:49:55     INFO -  REFTEST INFO | Saved log: Initializing canvas snapshot
12:49:55     INFO -  REFTEST INFO | Saved log: DoDrawWindow 0,0,800,1000
12:49:55     INFO -  REFTEST INFO | Saved log: [CONTENT] RecordResult fired
12:49:55     INFO -  REFTEST INFO | Saved log: RecordResult fired
Flags: needinfo?(mstange)
Whiteboard: [gfx-noted]
Do we have a diff of the preferences that change between Nightly and Aurora?
Flags: needinfo?(mstange) → needinfo?(ryanvm)
Not that I know of. Maybe go through https://dxr.mozilla.org/mozilla-central/source/mobile/android/app/mobile.js and https://dxr.mozilla.org/mozilla-central/source/modules/libpref/init/all.js for instances of NIGHTLY_BUILD? Assuming it's not something controlled at the ifdef level, anyway.

I'll try to bisect it down in the meantime too.
Flags: needinfo?(ryanvm)
Using my ghetto hacked-together script [1] I get these differences. There might be more.

43c43
< dom.select_events.enabled => false
---
> dom.select_events.enabled => true
190c190
< apz.record_checkerboarding => false
---
> apz.record_checkerboarding => true
387a388
> dom.input.dirpicker => true
422c423
< javascript.options.compact_on_user_inactive_delay => 300000
---
> javascript.options.compact_on_user_inactive_delay => 15000
437c438
< javascript.options.shared_memory => false
---
> javascript.options.shared_memory => true
923c924
< layout.css.float-logical-values.enabled => false
---
> layout.css.float-logical-values.enabled => true
935a937
> layout.css.unprefixing-service.globally-whitelisted => false
1224c1226
< signon.rememberSignons.visibilityToggle => false
---
> signon.rememberSignons.visibilityToggle => true
1282c1284
< webgl.enable-prototype-webgl2 => false
---
> webgl.enable-prototype-webgl2 => true
1305c1307
< layers.low-precision-buffer => true
---
> layers.low-precision-buffer => false
1537c1539
< browser.addon-watch.interval => -1
---
> browser.addon-watch.interval => 15000
1754c1756
< ui.zoomedview.enabled => false
---
> ui.zoomedview.enabled => true
1852a1855,1856
> layers.tiles.fade-in.enabled => true
> layers.tiles.fade-in.duration-ms => 250


[1] https://github.com/staktrace/moz-scripts/blob/master/allprefs.sh
Thanks! layers.low-precision-buffer sounds like the most probably cause here.
Yeah, especially considering the patch I landed in bug 1151071. Thought that one doesn't touch the content type of the layer itself, only of the tiles that back it. And the reftest shouldn't need low resolution tiles... hmm, not sure what's going on here after all.
Oh, I thought you were going to bisect through the prefs, not through the changesets. Still useful though, thanks.
I'm also realizing that this a Beta-only failure, so RELEASE_BUILD is probably the more relevant ifdef here.
Diff with and without RELEASE_BUILD on trunk m-c:

56c56
< dom.gamepad.non_standard_events.enabled => true
---
> dom.gamepad.non_standard_events.enabled => false
128c128
< media.getusermedia.screensharing.allowed_domains => mozilla.github.io,webex.com,*.webex.com,ciscospark.com,*.ciscospark.com,projectsquared.com,*.projectsquared.com,*.room.co,room.co,beta.talky.io,talky.io,*.clearslide.com,appear.in,*.appear.in,tokbox.com,*.tokbox.com,*.sso.francetelecom.fr,*.si.francetelecom.fr,*.sso.infra.ftgroup,*.multimedia-conference.orange-business.com,*.espacecollaboration.orange-business.com,free.gotomeeting.com,g2m.me,*.g2m.me,*.mypurecloud.com,*.mypurecloud.com.au,spreed.me,*.spreed.me,*.spreed.com,air.mozilla.org,*.circuit.com,*.yourcircuit.com,circuit.siemens.com,yourcircuit.siemens.com,circuitsandbox.net,*.unify.com,tandi.circuitsandbox.net,*.ericsson.net,*.cct.ericsson.net,*.opentok.com,*.conf.meetecho.com,meet.jit.si,*.meet.jit.si,web.stage.speakeasyapp.net,web.speakeasyapp.net,*.hipchat.me,*.beta-wspbx.com,*.wspbx.com,*.unifiedcloudit.com,*.smartboxuc.com,*.smartbox-uc.com,*.panterranetworks.com,pexipdemo.com,*.pexipdemo.com,pex.me,*.pex.me,*.rd.pexip.com,1click.io,*.1click.io,*.fuze.com,*.fuzemeeting.com,*.thinkingphones.com,gotomeeting.com,*.gotomeeting.com,gotowebinar.com,*.gotowebinar.com,gototraining.com,*.gototraining.com,citrix.com,*.citrix.com,expertcity.com,*.expertcity.com,citrixonline.com,*.citrixonline.com,g2m.me,*.g2m.me,gotomeet.me,*.gotomeet.me,gotomeet.at,*.gotomeet.at,miriadaxdes.miriadax.net,certificacion.miriadax.net,miriadax.net
---
> media.getusermedia.screensharing.allowed_domains => webex.com,*.webex.com,ciscospark.com,*.ciscospark.com,projectsquared.com,*.projectsquared.com,*.room.co,room.co,beta.talky.io,talky.io,*.clearslide.com,appear.in,*.appear.in,tokbox.com,*.tokbox.com,*.sso.francetelecom.fr,*.si.francetelecom.fr,*.sso.infra.ftgroup,*.multimedia-conference.orange-business.com,*.espacecollaboration.orange-business.com,free.gotomeeting.com,g2m.me,*.g2m.me,*.mypurecloud.com,*.mypurecloud.com.au,spreed.me,*.spreed.me,*.spreed.com,air.mozilla.org,*.circuit.com,*.yourcircuit.com,circuit.siemens.com,yourcircuit.siemens.com,circuitsandbox.net,*.unify.com,tandi.circuitsandbox.net,*.ericsson.net,*.cct.ericsson.net,*.opentok.com,*.conf.meetecho.com,meet.jit.si,*.meet.jit.si,web.stage.speakeasyapp.net,web.speakeasyapp.net,*.hipchat.me,*.beta-wspbx.com,*.wspbx.com,*.unifiedcloudit.com,*.smartboxuc.com,*.smartbox-uc.com,*.panterranetworks.com,pexipdemo.com,*.pexipdemo.com,pex.me,*.pex.me,*.rd.pexip.com,1click.io,*.1click.io,*.fuze.com,*.fuzemeeting.com,*.thinkingphones.com,gotomeeting.com,*.gotomeeting.com,gotowebinar.com,*.gotowebinar.com,gototraining.com,*.gototraining.com,citrix.com,*.citrix.com,expertcity.com,*.expertcity.com,citrixonline.com,*.citrixonline.com,g2m.me,*.g2m.me,gotomeet.me,*.gotomeet.me,gotomeet.at,*.gotomeet.at,miriadaxdes.miriadax.net,certificacion.miriadax.net,miriadax.net
512c512
< network.http.bypass-cachelock-threshold => 250
---
> network.http.bypass-cachelock-threshold => 200000
562c562
< network.jar.block-remote-files => true
---
> network.jar.block-remote-files => false
833,834c833,834
< services.blocklist.update_enabled => true
< security.onecrl.via.amo => false
---
> services.blocklist.update_enabled => false
> security.onecrl.via.amo => true
938,939c938,939
< layout.css.getBoxQuads.enabled => true
< layout.css.convertFromNode.enabled => true
---
> layout.css.getBoxQuads.enabled => false
> layout.css.convertFromNode.enabled => false
966c966
< layout.css.grid.enabled => true
---
> layout.css.grid.enabled => false
978c978
< layout.css.control-characters.visible => true
---
> layout.css.control-characters.visible => false
989c989
< dom.animations-api.core.enabled => true
---
> dom.animations-api.core.enabled => false
1018c1018
< dom.ipc.plugins.asyncdrawing.enabled => true
---
> dom.ipc.plugins.asyncdrawing.enabled => false
1026c1026
< svg.transform-box.enabled => true
---
> svg.transform-box.enabled => false
1301c1301
< webgl.enable-debug-renderer-info => true
---
> webgl.enable-debug-renderer-info => false
1359d1358
< dom.webnotifications.serviceworker.enabled => true
1363c1362
< full-screen-api.unprefix.enabled => true
---
> full-screen-api.unprefix.enabled => false
1427c1426
< dom.vr.enabled => true
---
> dom.vr.enabled => false
1460c1459
< dom.forms.inputmode => true
---
> dom.forms.inputmode => false
1606c1605
< dom.node.rootNode.enabled => true
---
> dom.node.rootNode.enabled => false
1719c1718
< browser.mirroring.enabled => true
---
> browser.mirroring.enabled => false
1889c1888
< consoleservice.logcat => true
---
> consoleservice.logcat => false
Try bisection confirms that this started when bug 1263192 landed.
This is hitting the Aurora simulations now, so I think we're back to comment 3 being relevant again :\
Summary: opaque-fractional-displayport reftests are going to permafail on Android when Gecko 49 merges to Beta → opaque-fractional-displayport reftests are going to permafail on Android when Gecko 49 merges to Aurora
See https://bugzilla.mozilla.org/show_bug.cgi?id=1270241#c4 - snorp's "no bug" patch to revert the layers.low-precision-buffer change got backed out because of these same reftest failures. So that's probably the relevant pref.
That is extremely useful information, thank you.
I don't care who does what to them, but getting rid of the permaorange blocks opening mozilla-aurora.
Severity: major → blocker
as philor mentioned this is a blocking reopening mozilla-aurora so could you please take a look at this so that we are able to reopen the trees, thanks!
Flags: needinfo?(mstange)
Sorry for losing track of this. I thought snorp's changeset to disable the test had landed in time to merge to Aurora. Has it not? I can disable the test on Android after lunch.

The test is a new test and not particularly important on Android, because lowres tiles are not opaque anyway. But since I don't completely understand why it fails, I didn't want to disable it earlier. But that time has passed now, and I still haven't got a chance to find out what happens, so we should disable the test on Android.
Snorp will disable the test. He didn't do it yesterday because inbound was closed.
Assignee: nobody → snorp
Flags: needinfo?(mstange)
Pushed by jwillcox@mozilla.com:
https://hg.mozilla.org/integration/mozilla-inbound/rev/c2d45532fcf4
Disable opaque-fractional-displayport reftests on Android r=mstange
https://hg.mozilla.org/mozilla-central/rev/c2d45532fcf4
Status: NEW → RESOLVED
Closed: 3 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla50
You need to log in before you can comment on or make changes to this bug.