2.90 KB, patch
|Details | Diff | Splinter Review|
14.86 KB, patch
|Details | Diff | Splinter Review|
Release Note Request (optional, but appreciated) [Why is this notable]: We're changing how canvas gets rendered on Linux. [Suggested wording]: Let's figure out the wording as it gets closer to being done. We should have information about performance at that point as well.
relnote-firefox: --- → ?
Created attachment 8733430 [details] [diff] [review] part 1 - backport zero-length gradient interval fix from Skia upstream This is just a straight backport of a fix I already had upstreamed but was not incorporated in time before Skia upstream tagged the m49 branch. It fixes some failing reftests.
Assignee: nobody → lsalzman
Status: NEW → ASSIGNED
Attachment #8733430 - Flags: review?(jmuizelaar)
Created attachment 8733432 [details] [diff] [review] part 2 - enable Skia canvas on Linux and fix web-platform-tests/reftests The main effect of this patch is to pref on Skia canvas on Linux. It removes a lot of meta-data we had for web-platform-test fails with Cairo. Since now Linux is not a special snowflake on canvas results, using Skia like everything else, a lot of "unexpected-pass"es showed up, so this removes those. Adds one actual WPT failure in Linux that matches existing failures on Mac and Windows. Another reftest that is insane and actually skipped on all platforms but Linux (drawCustomFocusRing.html) had to be fuzzed due to 4 corner pixels blending weirdly from inside an onFocus handler.
Attachment #8733432 - Flags: review?(jmuizelaar)
Try run: https://treeherder.mozilla.org/#/jobs?repo=try&revision=91b8128f7267 Everything looks green, ignoring the current intermittent reftest flu that is currently going around on some talos results.
(In reply to Lee Salzman [:lsalzman] from comment #4) > Try run: https://treeherder.mozilla.org/#/jobs?repo=try&revision=91b8128f7267 > > Everything looks green, ignoring the current intermittent reftest flu that > is currently going around on some talos results. It should also be noted that the talos results are more or less unchanged. Bigger differences we might have previously noticed were addressed when we disabled xrender by default with Cairo still in place. So if nothing else, we at least don't regress and now our canvas results should be more or less consistent across all platforms.
Attachment #8733430 - Flags: review?(jmuizelaar) → review+
Attachment #8733432 - Flags: review?(jmuizelaar) → review+
https://hg.mozilla.org/mozilla-central/rev/a37ac3dbec9f https://hg.mozilla.org/mozilla-central/rev/180f3f0494a7 https://hg.mozilla.org/mozilla-central/rev/826afa1625f4
Status: ASSIGNED → RESOLVED
Last Resolved: 3 years ago
status-firefox48: --- → fixed
Resolution: --- → FIXED
Target Milestone: --- → mozilla48
We're going to let go with 48 to dev edition, but if we encounter any blocking problems, may keep it twice on dev edition.
Added to the release notes with "Skia enabled by default on GNU/Linux" as wording
relnote-firefox: ? → 48+
Milan, can we have more information or a blog post? For now "Skia enabled by default on GNU/Linux" is cryptic, even for geek like us...
"Skia, the graphic library, enabled by default on GNU/Linux for better performances" => I updated to that
"The use of Skia, the graphic library, enabled by default on GNU/Linux for better canvas performance." It is just canvas, so it's worth mentioning that.
You need to log in before you can comment on or make changes to this bug.