1611692 - 72.0.2 rendering totally broken MacBook (Retina, 12-inch, Early 2016) running OSX 10.11.6. Can't be used

Status: NEW

(Core :: Graphics, defect, P3)

Description

[Lauri Raittila]

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:70.0) Gecko/20100101 Firefox/70.0

Steps to reproduce:

Automatic update after booting the machine.

Actual results:

Everything just jumps around on screen. Apparently everything works under the hood, but it is imnpossible to use at all

Expected results:

Normal

Comment 1

[Lauri Raittila]

Same problem also on Firefox Developer edition. See attacment for screenshot

Comment 2

[Lauri Raittila]

Attachment: screenshot of problem

Comment 3

[Lauri Raittila]

And 72.0.1 was working fine

Comment 4

[Sotaro Ikeda [:sotaro]]

Lauri Raittila, can you upload about:support to this bug?

Comment 5

[Sotaro Ikeda [:sotaro]]

:mstange, do you have an idea about this bug?

Comment 6

[Lauri Raittila]

Attachment: about:support on developer version.

(In reply to Sotaro Ikeda [:sotaro] from comment #4)
> Lauri Raittila, can you upload about:support to this bug?

Comment 7

[Lauri Raittila]

Similar problem in both 72.0.2 and 73.0b9. Already replaced 72.0.2 with earlier version.

Comment 8

[Markus Stange [:mstange]]

Intel(R) HD Graphics 515
Vendor ID: 0x8086
Device ID: 0x191e

Comment 9

[Markus Stange [:mstange]]

Lauri, does this appear as soon as Firefox starts, or are things fine for a while and then break later?

Comment 10

[Lauri Raittila]

Immidiately after Firefox starts.

Comment 11

[Markus Stange [:mstange]]

Thanks. I will try to find similar hardware and try to reproduce this problem.

Comment 12

[Markus Stange [:mstange]]

In the meantime, as a workaround, it should be possible to start the browser in safe mode, go to the preferences, and then disable hardware acceleration in the Performance section. (Uncheck the "Use recommended performance settings" checkbox first.)

Comment 13

[Lauri Raittila]

Workaround works on 72.0.2. Thanks.

Comment 14

[Sotaro Ikeda [:sotaro]]

On chromium, the gpu rendering is disabled and software rendering is used.
https://cs.chromium.org/chromium/src/gpu/config/software_rendering_list.json?type=cs&q=0x191e&sq=package:chromium&g=0&l=972

Comment 15

[Markus Stange [:mstange]]

Hi Lauri, I'm curious whether this bug is fixed in macOS 10.12 or 10.13. Could you try to update your macOS to a newer version and test again? Thanks!

Jeff and I have been looking for hardware that reproduces this. We think it's rather rare: It needs to have a Skylake CPU, and it needs to have been released before September 2016 because it would have needed to ship with macOS 10.11. We think it's really just that one Macbook model ("Early 2016") that is affected.

Comment 16

[Markus Stange [:mstange]]

You can use this link to get macOS 10.12 (High Sierra) even though Apple will usually try to get you onto something more recent: https://itunes.apple.com/us/app/macos-high-sierra/id1246284741?ls=1&mt=12 (from https://support.apple.com/en-ca/HT208969)

Comment 18

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Bug 1611692 - Block OSX 10.11 and Intel HD 515 from hardware acceleration. r=jrmuizel

Comment 19

[Jeff Muizelaar [:jrmuizel]]

Some documentation for only blocking 191e:

https://everymac.com/ultimate-mac-lookup/?search_keywords=A1534 lists 3 different 2016 MacBooks

• https://everymac.com/systems/apple/macbook/specs/macbook-core-m3-1.1-12-early-2016-specs.html
• https://everymac.com/systems/apple/macbook/specs/macbook-core-m5-1.2-12-early-2016-specs.html
• https://everymac.com/systems/apple/macbook/specs/macbook-core-m7-1.3-12-early-2016-specs.html

All of these are listed as having a HD Graphics 515

https://en.wikipedia.org/wiki/List_of_Intel_graphics_processing_units#Gen9 gives the same 191e id for all processors.

Comment 20

[Markus Stange [:mstange]]

Telemetry indicates that uptake of Firefox 72 was a lot lower than previous Firefox versions on the affected machines. Initially I suspected that this bug started happening with 70 (which was when CoreAnimation support shipped), but 70 does not show as strong a change as 72 in the numbers we were looking at. We don't really have an explanation for this yet.
One change that shipped in 72 was bug 1592150, which caused us to use BlitFramebuffer between the different swap chain IOSurface buffers of one NativeLayerCA.

Comment 21

[BugBot [:suhaib / :marco/ :calixte]]

Clear a needinfo that is pending on an inactive user.

Inactive users most likely will not respond; if the missing information is essential and cannot be collected another way, the bug maybe should be closed as INCOMPLETE.

For more information, please visit BugBot documentation.