1554777 - Intermittent Tier 2 /css/css-backgrounds/border-image-width-006.xht | Testing http://web-platform.test:8000/css/css-backgrounds/border-image-width-006.xht == http://web-platform.test:8000/css/reference/ref-filled-green-100px-square.xht

Status: RESOLVED FIXED

(Core :: Layout: Images, Video, and HTML Frames, defect, P5)

Description

[Treeherder Bug Filer]

Filed by: rgurzau [at] mozilla.com
Parsed log: https://treeherder.mozilla.org/logviewer.html#?job_id=248555784&repo=mozilla-central
Full log: https://queue.taskcluster.net/v1/task/Ghb5BDfITxiOcKJ3dpIaCw/runs/0/artifacts/public/logs/live_backing.log
Reftest URL: https://hg.mozilla.org/mozilla-central/raw-file/tip/layout/tools/reftest/reftest-analyzer.xhtml#logurl=https://queue.taskcluster.net/v1/task/Ghb5BDfITxiOcKJ3dpIaCw/runs/0/artifacts/public/logs/live_backing.log&only_show_unexpected=1

---

[task 2019-05-27T15:29:23.375Z] 15:29:23 INFO - TEST-START | /css/css-backgrounds/border-image-width-006.xht
[task 2019-05-27T15:29:23.888Z] 15:29:23 INFO - TEST-UNEXPECTED-FAIL | /css/css-backgrounds/border-image-width-006.xht | Testing http://web-platform.test:8000/css/css-backgrounds/border-image-width-006.xht == http://web-platform.test:8000/css/reference/ref-filled-green-100px-square.xht
[task 2019-05-27T15:29:23.889Z] 15:29:23 INFO - REFTEST

Comment 1

[Intermittent Failures Robot]

4 failures in 4337 pushes (0.001 failures/push) were associated with this bug in the last 7 days.

Repository breakdown:

• mozilla-central: 4

Platform breakdown:

• android-em-7-0-x86_64: 4

For more details, see:
https://treeherder.mozilla.org/intermittent-failures.html#/bugdetails?bug=1554777&startday=2019-05-27&endday=2019-06-02&tree=all

Comment 2

[Intermittent Failures Robot]

5 failures in 5197 pushes (0.001 failures/push) were associated with this bug in the last 7 days.

Repository breakdown:

• try: 5

Platform breakdown:

• android-em-7-0-x86_64: 5

For more details, see:
https://treeherder.mozilla.org/intermittent-failures.html#/bugdetails?bug=1554777&startday=2019-06-03&endday=2019-06-09&tree=all

Comment 3

[Intermittent Failures Robot]

1 failures in 4753 pushes (0.0 failures/push) were associated with this bug in the last 7 days.

Repository breakdown:

• mozilla-central: 1

Platform breakdown:

• android-em-7-0-x86_64: 1

For more details, see:
https://treeherder.mozilla.org/intermittent-failures.html#/bugdetails?bug=1554777&startday=2019-06-10&endday=2019-06-16&tree=all

Comment 4

[Cosmin Sabou [:CosminS]]

Looks like this started failing from: https://treeherder.mozilla.org/#/jobs?repo=autoland&searchStr=android%2C7.0%2Cx86-64%2Copt%2Cweb%2Cplatform%2Ctests%2Ctest-android-em-7.0-x86_64%2Fopt-web-platform-tests-reftests-e10s-1%2Cw%28wr1%29&tochange=253fa3bd2d74f4f0aa694f84ccfec2d2a93849b7&fromchange=6aa8a1cb9c5cd5d957ad019acf3e665aa0e546d2&group_state=expanded&selectedJob=252720655

https://hg.mozilla.org/integration/autoland/rev/59ed9ee0bd98a2afb32a068186ce174af3adc8f0

Comment 5

[Maja Frydrychowicz :impossibus (was :maja_zf) (needinfo me)]

It looks like an intermittent that started last month when we turned on these tests for GeckoView. It needs investigation from the appropriate team (Layout?). Feel free to update the metadata if it becomes a problem, but for now it doesn't seem that frequent and doesn't warrant disabling in my opinion.

Comment 6

[Intermittent Failures Robot]

24 failures in 562 pushes (0.043 failures/push) were associated with this bug yesterday.

Repository breakdown:

• autoland: 19
• mozilla-inbound: 2
• try: 2
• mozilla-central: 1

Platform breakdown:

• android-em-7-0-x86_64: 24

For more details, see:
https://treeherder.mozilla.org/intermittent-failures.html#/bugdetails?bug=1554777&startday=2019-06-20&endday=2019-06-20&tree=all

Comment 7

[Intermittent Failures Robot]

38 failures in 2581 pushes (0.015 failures/push) were associated with this bug in the last 7 days.

** This failure happened more than 30 times this week! Resolving this bug is a high priority. **

** Try to resolve this bug as soon as possible. If unresolved for 2 weeks, the affected test(s) may be disabled. **

Repository breakdown:

• autoland: 27
• try: 4
• mozilla-inbound: 3
• mozilla-central: 4

Platform breakdown:

• android-em-7-0-x86_64: 38

For more details, see:
https://treeherder.mozilla.org/intermittent-failures.html#/bugdetails?bug=1554777&startday=2019-06-17&endday=2019-06-23&tree=all

Comment 8

[Intermittent Failures Robot]

15 failures in 847 pushes (0.018 failures/push) were associated with this bug yesterday.

Repository breakdown:

• autoland: 10
• try: 1
• mozilla-central: 1
• mozilla-inbound: 3

Platform breakdown:

• android-em-7-0-x86_64: 15

For more details, see:
https://treeherder.mozilla.org/intermittent-failures.html#/bugdetails?bug=1554777&startday=2019-06-25&endday=2019-06-25&tree=all

Comment 9

[Intermittent Failures Robot]

20 failures in 844 pushes (0.024 failures/push) were associated with this bug yesterday.

Repository breakdown:

• try: 2
• autoland: 9
• mozilla-central: 6
• mozilla-inbound: 3

Platform breakdown:

• android-em-7-0-x86_64: 20

For more details, see:
https://treeherder.mozilla.org/intermittent-failures.html#/bugdetails?bug=1554777&startday=2019-06-26&endday=2019-06-26&tree=all

Comment 10

[Andreea Pavel [:apavel]]

Attachment: Bug 1554777 - update test expectation r=jmaher

Comment 11

[Intermittent Failures Robot]

35 failures in 792 pushes (0.044 failures/push) were associated with this bug yesterday.

Repository breakdown:

• mozilla-inbound: 6
• autoland: 26
• mozilla-central: 3

Platform breakdown:

• android-em-7-0-x86_64: 35

For more details, see:
https://treeherder.mozilla.org/intermittent-failures.html#/bugdetails?bug=1554777&startday=2019-06-27&endday=2019-06-27&tree=all

Comment 12

[Maja Frydrychowicz :impossibus (was :maja_zf) (needinfo me)]

Attachment: logcat of repeated manual runs

This test should show a green box. If it fails, it shows a red box. It depends on an image resource.

I can reproduce the intermittent failure outside of the wpt harness by running:

 adb shell am start -W -n org.mozilla.geckoview.test/org.mozilla.geckoview.test.TestRunnerActivity -a android.intent.action.MAIN -es args "http://web-platform.test:8000/css/css-backgrounds/border-image-width-006.xht"

Every 5-10 runs I see that the green box doesn't appear at all until I click/touch the window. When it fails, the page load takes a bit longer and emulator screen flickers black. The image in the test always loads fine, so that's not the issue. Looking at logcat, I think it might be related to Android resource management and/or activity focus? Maybe we lose focus?

In the attached log, you can see that when the green box does not appear (FAIL)
there messages below appear before we get a pageshow event:

• WindowManager: Destroying surface...
• dispatch GeckoView:SetActive, data={"active":false}
• Choreographer: Skipped 66 frames! The application may be doing too much work on its main thread

In a failing run, the last logcat messages shown once we see a red square are:

06-28 21:19:37.770  2938  2953 D GeckoViewContent[C]: handleEvent: pageshow
06-28 21:19:38.132  2877  2892 W GeckoEventDispatcher: No listener for GeckoView:StateUpdated

and then when I click/touch the window, green square appears and I get:

06-28 21:20:45.475  2877  2922 D EGL_emulation: eglMakeCurrent: 0xc6d9fa60: ver 2 0 (tinfo 0xd1fcbdf0)

I'm not sure why this test in particular is affected. Presumably this kind of situation occurs all the time on Android...

Comment 13

[Intermittent Failures Robot]

37 failures in 815 pushes (0.045 failures/push) were associated with this bug yesterday.

** This test has failed more than 150 times in the last 21 days. It should be disabled until it can be fixed. **

Repository breakdown:

• try: 3
• autoland: 22
• mozilla-inbound: 10
• mozilla-central: 2

Platform breakdown:

• android-em-7-0-x86_64: 37

For more details, see:
https://treeherder.mozilla.org/intermittent-failures.html#/bugdetails?bug=1554777&startday=2019-06-28&endday=2019-06-28&tree=all

Comment 14

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/48d35c2a2dd5
update test expectation r=jmaher

Comment 15

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/d34e20b62877
Disable test on Android because of flakiness and frequent failures. a=testonly

Comment 16

[Cosmin Sabou [:CosminS]]

I've disabled this because of it's flakiness. On the push that Andreea updated the expectations it started failing as TEST-UNEXPECTED-PASS https://treeherder.mozilla.org/#/jobs?repo=autoland&selectedJob=254152507&resultStatus=testfailed%2Cbusted%2Cexception&revision=48d35c2a2dd5cd8301cee595c0238366d67a34c4&searchStr=wr%2Candroid

Comment 17

[Arthur Iakab [arthur_iakab]]

https://hg.mozilla.org/mozilla-central/rev/48d35c2a2dd5
https://hg.mozilla.org/mozilla-central/rev/d34e20b62877

Comment 18

[Intermittent Failures Robot]

129 failures in 4409 pushes (0.029 failures/push) were associated with this bug in the last 7 days.

This is the #7 most frequent failure this week.

** This failure happened more than 75 times this week! Resolving this bug is a very high priority. **

** Try to resolve this bug as soon as possible. If unresolved for 1 week, the affected test(s) may be disabled. **

Repository breakdown:

• autoland: 81
• mozilla-central: 14
• try: 10
• mozilla-inbound: 24

Platform breakdown:

• android-em-7-0-x86_64: 129

For more details, see:
https://treeherder.mozilla.org/intermittent-failures.html#/bugdetails?bug=1554777&startday=2019-06-24&endday=2019-06-30&tree=all

Comment 19

[Daniel Holbert [:dholbert]]

As Maja observed in bug 1567639, this test (viewable at http://web-platform-tests.live/css/css-backgrounds/border-image-width-006.xht ) often fails on first load even in Firefox Nightly on Desktop, until you click or interact with the page.

Comment 20

[Daniel Holbert [:dholbert]]

I suspect this may have to do with the our image decoding -- not sure whether it's the off-main-thread-decoding or some other aspect of lazy image decoding, but at least if I toggle the pref image.decode-immediately.enabled to true then I can't reproduce anymore (on Desktop).

Comment 21

[Daniel Holbert [:dholbert]]

emilio, heycam mentioned that you'd worked on a similar issue where image async-decode wasn't working properly with onload-unblocking... Did you run into anything like this where images weren't getting painted/invalidated when decoded?

Comment 22

[Daniel Holbert [:dholbert]]

FWIW here's a profile of me loading the web-hosted testcase twice, in desktop Nightly. The bug does not reproduce the first time, but it does reproduce the second time.
Profile: https://perfht.ml/2TaRbEQ

On the first load, the image shows up on its own a little after 5 seconds. Then I shift+reloaded, and the image never showed up on its own (i.e. the bug reproduced!) though it did show up when I clicked the page at around 15 seconds.

Comment 23

[Timothy Nikkel (:tnikkel)]

This doesn't reproduce on mac for me, but is really easy to reproduce on linux.

I'm guessing the problem has something to do with calling StartDecoding during the paint here https://searchfox.org/mozilla-central/rev/e0b0c38ee83f99d3cf868bad525ace4a395039f1/layout/painting/nsImageRenderer.cpp#110 and then painting the image, and then delivering the notifications for the decode being finishing after the paint. My guess would be that when the notifications arrive the receiver probably thinks it already painted and so ignores them? But we need to invalidate so the pixels make it to the screen and not just layer contents.

Comment 24

[Emilio Cobos Álvarez (:emilio)]

This only repros with WebRender disabled.

I can try to take a look.

Comment 25

[Emilio Cobos Álvarez (:emilio)]

Attachment: Bug 1554777 - Call SchedulePaint() rather than MarkNeedsDisplayItemRebuild() when we get the size available notification for a style image. r=mattwoodrow,tnikkel

So the issue here is that this test-case has a zero-sized border, but still with
a border-image (which we should paint).

So the first time we get here, the image is not loaded, and thus we don't get
here:

https://searchfox.org/mozilla-central/rev/e0b0c38ee83f99d3cf868bad525ace4a395039f1/layout/painting/nsDisplayList.h#4254

Which means that we end up with zero bounds and thus we don't even get to the
border painting code.

However, when the image loads, we get to MarkNeedsDisplayItemRebuild(), but that
doesn't schedule any paint, only marks the frames as modified in order for
display items to be rebuilt eventually.

Thus eventually, where we force a repaint by other means, we paint correctly.

This only works in more general cases because we get to the nsImageRenderer code
which does vastly different stuff.

InvalidateFrame() seems to do the right thing and schedule a paint, so use it.
It's not clear to me which one of nsIFrame::InvalidateFrame() or
nsIFrame::SchedulePaint() we should use...

If I understand correctly, InvalidateFrame() will only do something iff there
are display items for the frame, so that should make the IsVisible() check
redundant.

Note however that I think there's still a race condition, if we get to paint in
between the SIZE_AVAILABLE notification (the one where we actually invalidate
the display items), and the LOAD notification (the one the border-image code
checks).

I'll send a separate patch for that, I think SIZE_AVAILABLE should be a
strong-enough hint and that allows us to remove nsStyleImage::IsLoaded()...

The RequestReflow stuff also looks highly suspicious... shape-outside
sync-decodes, and it seems we could end up invalidating reflow from the reflow
code...

Comment 26

[Emilio Cobos Álvarez (:emilio)]

Attachment: Bug 1554777 - Use SIZE_AVAILABLE rather than loaded to figure out whether to try to draw a border image.

The current code needs to handle incomplete draws already due to async decoding,
and this prevents the race condition where we paint between the size-available
and load notifications, as the CSS image loader only invalidates for the first.

Once we try to decode the image at least once, then we'd repaint properly from
ImageLoader::OnFrameComplete.

Comment 27

[Emilio Cobos Álvarez (:emilio)]

Attachment: Bug 1554777 - Re-enable test.

I think they should pass after the last few patches.

Comment 28

[Daniel Holbert [:dholbert]]

Thanks, Emilio!

Comment 29

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/da9ae31b5000
Call SchedulePaint() rather than MarkNeedsDisplayItemRebuild() when we get the size available notification for a style image. r=mattwoodrow,tnikkel
https://hg.mozilla.org/integration/autoland/rev/87ed773c0a80
Use SIZE_AVAILABLE rather than loaded to figure out whether to try to draw a border image. r=tnikkel

Comment 30

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/9cd5f82e915c
Re-enable test. r=dholbert

Comment 31

[Cristina Coroiu [:ccoroiu]]

https://hg.mozilla.org/mozilla-central/rev/da9ae31b5000
https://hg.mozilla.org/mozilla-central/rev/87ed773c0a80
https://hg.mozilla.org/mozilla-central/rev/9cd5f82e915c