1667267 - Scaling problem with Flash content and native Macbook display on Firefox 81.0

Status: VERIFIED FIXED

(Core Graveyard :: Plug-ins, defect, P1)

Description

[Jeromie Clark]

Attachment: Actual Ouptut in Firefox 81.0

I attempted to bisect this with mozregression, but I can't identify a corresponding change. This only appears to happen with the generally available release version of Firefox 81.0. It does not happen on Firefox 82 or nightly. I'm mostly concerned that this was a late breaking change that might get merged into another branch in the future.

To Reproduce:
Using the native retina display on a relatively recent Macbook
In my case, macOS 10.15.6 with Firefox 81.0 and Flash Player 32.0.0.433 with a 2019 16" Macbook Pro.
Launch Firefox 81.0 (64-bit)
Go to: http://www.homestarrunner.com/
Allow Flash to run

Actual:
Flash content is scaled way up to the point that it's unusable (see screenshot)

Expected:
Flash content rendered at the correct scale

Comment 1

[Jeromie Clark]

Attachment: Expected Output as seen in Firefox 82.0 Nightly

Comment 2

[Jeromie Clark]

Attachment: Attempted bisect, but couldn't find a prior breaking build

Comment 3

[[:philipp]]

we're also getting this problem reported on support.mozilla.org a couple of times:
https://support.mozilla.org/en-US/questions/firefox?tagged=bug1667267&show=all&owner=all

Comment 4

[Jeromie Clark]

@cpeterson - Per our discussion, I attempted to regress for the fix build, but I never actually see the bug reproduce in the range between 81 and 82, presumably because I'm getting Nightly builds and not the ones from the generally available branch. Is there a flag/setting that I'm missing that would get me builds from the correct branch? I attached logs from that attempt. Hopefully it's at least useful for an exclusionary range. :D

Comment 5

[Jeromie Clark]

Attachment: Attempt to bisect for a build with a fix, also unsuccessful

Comment 6

[Chris Peterson [:cpeterson]]

Is there a flag/setting that I'm missing that would get me builds from the correct branch? I attached logs from that attempt. Hopefully it's at least useful for an exclusionary range. :D

You can tell mozregression to download builds from the Beta or Release channels using the flag --repo mozilla-beta or --repo mozilla-release. But bisecting the Release channel is typically not helpful because there are so few commits and builds.

btw, I recommend the mozregression GUI app. It's easier to configure and run.

https://mozilla.github.io/mozregression/install.html

I can reproduce this bug with Flash Player 32.0.0.433 in Firefox 81.0 on macOS 10.15.6. I had a hunch why we can reproduce this bug on the Release channel and not Nightly: WebRender is enabled on Nightly but not Beta or Release yet. This regression is in the non-WebRender graphics code. I will bisect Nightly with WebRender turned off.

Comment 7

[Chris Peterson [:cpeterson]]

[Tracking Requested - why for this release]:

Ryan and Matt, Firefox 81 breaks Flash rendering and is probably worth fixing in an 81.0.x dot release. I believe this is a regression from scaling bug 1653166. That bug has at least five other regressions.

I bisected this regression to your pushlog here, which includes scaling bug 1653166:

https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=4ed6ec0c8220321b13c108ffaa7ab6cfc8d7698f&tochange=4558b3cfd6fe49647780a0ec86dfa97f0dd3adcb

Comment 8

[Ryan VanderMeulen [:RyanVM]]

Would be great if we understood why this doesn't reproduce on 82+. Or does it with WR disabled? Also, are we sure this only affects macOS?

Comment 9

[Chris Peterson [:cpeterson]]

(In reply to Ryan VanderMeulen [:RyanVM] from comment #8)

Would be great if we understood why this doesn't reproduce on 82+. Or does it with WR disabled? Also, are we sure this only affects macOS?

I can reproduce this bug on 82+ if I disable WR (with gfx.webrender.force-disabled pref = true). I tested both macOS and Windows without WebRender and the problem only affects macOS.

Comment 10

[Ryan VanderMeulen [:RyanVM]]

Thanks, this is sounding more and more like a dot release driver.

Comment 11

[Jeromie Clark]

We don't have access to our offices or labs. The only high density display I have access to at the moment is the native macbook display. I can't authoritatively rule out Windows based on our testing. We just happened to stumble across it.

Comment 12

[Matt Woodrow (:mattwoodrow)]

Attachment: Bug 1667267 - Set MacIOSurfaceImage picture rect in device pixels. r?mstange

Comment 13

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/5853ab33adb8
Set MacIOSurfaceImage picture rect in device pixels. r=mstange

Comment 14

[Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)]

https://hg.mozilla.org/mozilla-central/rev/5853ab33adb8

Comment 15

[Ryan VanderMeulen [:RyanVM]]

Please also request Beta and Release approval on this patch ASAP.

Comment 16

[Matt Woodrow (:mattwoodrow)]

Comment on attachment 9178136 [details]
Bug 1667267 - Set MacIOSurfaceImage picture rect in device pixels. r?mstange

Beta/Release Uplift Approval Request

• User impact if declined: Incorrectly scaled flash content on HiDPI monitors on MacOS.
• Is this code covered by automated tests?: No
• Has the fix been verified in Nightly?: Yes
• Needs manual test from QE?: No
• If yes, steps to reproduce:
• List of other uplifts needed: None
• Risk to taking this patch: Low
• Why is the change risky/not risky? (and alternatives if risky): Simple fix to take the device scaling into account when computing plugin image bounds
• String changes made/needed: None

Comment 17

[Julien Cristau [:jcristau] (back April 22)]

Comment on attachment 9178136 [details]
Bug 1667267 - Set MacIOSurfaceImage picture rect in device pixels. r?mstange

regression fix, approved for 82.0b5

Comment 18

[Julien Cristau [:jcristau] (back April 22)]

https://hg.mozilla.org/releases/mozilla-beta/rev/b901ddcb94e4

Comment 19

[Ryan VanderMeulen [:RyanVM]]

Comment on attachment 9178136 [details]
Bug 1667267 - Set MacIOSurfaceImage picture rect in device pixels. r?mstange

Approved for 81.0.1.

Comment 20

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-release/rev/13b28fd6ed10

Comment 21

[Ryan VanderMeulen [:RyanVM]]

Added to the Firefox 81.0.1 relnotes:

Resolved incorrect scaling of Flash content on HiDPI macOS systems

Comment 22

[Timea Cernea [:tbabos][inactive]]

Reproduced the issue on affected Release 81.0 on MacOS 10.15 MacBook Pro (13inch, 2017, 2560 x 1600).
Verified-fixed on all latest Firefox Nightly 83.0a1 (2020-09-30), Beta 85.0b5 and Release 80.0.1. on the same macOS device.