Add a reftest-resolution attribute for painting with a specified resolution
Categories
(Testing :: Reftest, enhancement)
Tracking
(firefox69 fixed)
| Tracking | Status | |
|---|---|---|
| firefox69 | --- | fixed |
People
(Reporter: botond, Assigned: botond)
References
Details
Attachments
(3 files)
We have two reftest attributes currently related to zooming:
reftest-zoomsets a full-zoom on the pagereftest-async-zoomset a resolution, but only in the compositor, thus exercising the codepath where the user has pinch-zoomed and we've reacted to that in the compostor, but have not yet repainted at the new resolution
We also want to be able to test page rendering after the page has been repainted at the new resolution after a pinch-zoom. We currently test this by using <meta name="viewport" content="initial-scale=..."> which causes us to paint the page at the specified scale.
However, processing a meta viewport tag has side effects beyond setting a resolution: it triggers the mobile viewport sizing logic added in bug 1423013 and relatives.
For desktop zooming, we'd like to be able to test painting with a resolution without triggering the mobile viewport sizing logic. (For at least two reasons: (1) we want to be testing configurations that are as close as possible to what happens in production; and (2) the mobile viewport sizing logic has interactions with layout scrollbars and I'd rather not support the combination if I don't have to.)
To this end, I'd like to add a reftest-resolution attribute that's similar to reftest-async-zoom but applies the resolution during painting rather than just during composition.
| Assignee | ||
Comment 1•6 years ago
|
||
| Assignee | ||
Comment 2•6 years ago
|
||
With this change, the tests in question pass on desktop, except for
scrollbar-zoom-resolution-2.html which is annotated as failing.
Depends on D32774
Comment 3•6 years ago
|
||
Bugbug thinks this bug is a enhancement, but please change it back in case of error.
Updated•6 years ago
|
Comment 5•6 years ago
|
||
Backed out 2 changesets (Bug 1554790) for failures in reftest-resolution.html
Failure log: https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=250215196&repo=autoland&lineNumber=62845
Backout: https://hg.mozilla.org/integration/autoland/rev/06f8a3a6c43f3d1902b72d8e31f446d88f7bd3f8
Comment 6•6 years ago
|
||
Another relevant fail example in the above backout:
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=250215217&repo=autoland&lineNumber=35621
| Assignee | ||
Updated•6 years ago
|
| Assignee | ||
Comment 7•6 years ago
|
||
(In reply to Noemi Erli[:noemi_erli] from comment #6)
Another relevant fail example in the above backout:
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=250215217&repo=autoland&lineNumber=35621
That Wr4 failure seems to be pre-existing, as it occurs without these patches too: https://treeherder.mozilla.org/#/jobs?repo=try&revision=94bc43eeb4b013286ad07e2c3a5bbdd84970927e
| Assignee | ||
Comment 8•6 years ago
|
||
After a lot of debugging I tracked down the cause of Windows reftest failure. It's related to advanced layers: a codepath implemented for ContainerLayerComposite needed to be put in place for ContainerLayerMLGPU as well.
New Try push: https://treeherder.mozilla.org/#/jobs?repo=try&revision=b10e0f52b91e0bbddb8ba2df077f081da4e341c6
| Assignee | ||
Comment 9•6 years ago
|
||
Comment 10•6 years ago
|
||
Comment 11•6 years ago
|
||
| bugherder | ||
https://hg.mozilla.org/mozilla-central/rev/b418fcf242aa
https://hg.mozilla.org/mozilla-central/rev/421c96010e67
https://hg.mozilla.org/mozilla-central/rev/4a1c99839cec
| Assignee | ||
Comment 12•6 years ago
|
||
I didn't realize this until now, but these patches (specifically D34422) also fixed a bug that made pinch-zooming on Windows basically unusable because the resolution at which the content is rendered would jump back and forth. \o/
Description
•