1545859 - set gfx.font_ahem_antialias_none=true for all reftests

Status: RESOLVED FIXED

(Core :: Layout: Text and Fonts, task, P3)

Description

[Joel Maher ( :jmaher ) (UTC -0800)]

it appears that by doing this we have some visible changes in windows and osx reftests. Most of these are passing conditions which is great.

There is one issue on windows 7 debug where I see a crash:
https://treeherder.mozilla.org/#/jobs?repo=try&searchStr=win%2Creftest&revision=5871ae98f70b92f1761e12af59acd2c478ebf970

it looks to start with an assertion:

Comment 1

[Joel Maher ( :jmaher ) (UTC -0800)]

the assertion is here:
20:21:56 INFO - Assertion failure: false (d3dcompiler DLL loading failed.), at z:/build/build/src/gfx/gl/GLLibraryEGL.cpp:431
20:22:27 INFO - #01: mozilla::gl::GLLibraryEGL::EnsureInitialized(bool,nsTSubstring<char> * const) [gfx/gl/GLLibraryEGL.cpp:385]
20:22:27 INFO -
20:22:27 INFO - #02: mozilla::gl::GLContextProviderEGL::CreateOffscreen(mozilla::gfx::IntSizeTyped<mozilla::gfx::UnknownUnits> const &,mozilla::gl::SurfaceCaps const &,mozilla::gl::CreateContextFlags,nsTSubstring<char> * const) [gfx/gl/GLContextProviderEGL.cpp:1130]
20:22:27 INFO -
20:22:27 INFO - #03: mozilla::WebGLContext::CreateAndInitGL(bool,std::vector<mozilla::WebGLContext::FailureReason,std::allocator<mozilla::WebGLContext::FailureReason> > * const) [dom/canvas/WebGLContext.cpp:590]
20:22:27 INFO -
20:22:27 INFO - #04: mozilla::WebGLContext::SetDimensions(int,int) [dom/canvas/WebGLContext.cpp:843]
20:22:27 INFO -

and then ends up with a crash:
20:22:51 INFO - Thread 0 (crashed)
20:22:51 INFO - 0 xul.dll!mozilla::gl::GLLibraryEGL::DoEnsureInitialized(bool,nsTSubstring<char> * const) [GLLibraryEGL.cpp:0160424142d14988f7b30595e03f156c31278a42 : 431 + 0x0]
20:22:51 INFO - eip = 0x524e5977 esp = 0x0032d368 ebp = 0x0032d7c4 ebx = 0x00000000
20:22:51 INFO - esi = 0x00e97890 edi = 0x0032d484 eax = 0x739be6b0 ecx = 0x000001af
20:22:51 INFO - edx = 0x00000049 efl = 0x00000202
20:22:51 INFO - Found by: given as instruction pointer in context
20:22:51 INFO - 1 xul.dll!mozilla::gl::GLLibraryEGL::EnsureInitialized(bool,nsTSubstring<char> * const) [GLLibraryEGL.cpp:0160424142d14988f7b30595e03f156c31278a42 : 385 + 0x11]
20:22:51 INFO - eip = 0x524dc746 esp = 0x0032d7cc ebp = 0x0032d7d8 ebx = 0x0c530173
20:22:51 INFO - esi = 0x00e97890 edi = 0x0c53a800
20:22:51 INFO - Found by: call frame info
20:22:51 INFO - 2 xul.dll!mozilla::gl::GLContextProviderEGL::CreateOffscreen(mozilla::gfx::IntSizeTyped<mozilla::gfx::UnknownUnits> const &,mozilla::gl::SurfaceCaps const &,mozilla::gl::CreateContextFlags,nsTSubstring<char> * const) [GLContextProviderEGL.cpp:0160424142d14988f7b30595e03f156c31278a42 : 1130 + 0xc]
20:22:51 INFO - eip = 0x524e0bfb esp = 0x0032d7e0 ebp = 0x0032d824 esi = 0x0032d858
20:22:51 INFO - Found by: call frame info
20:22:51 INFO - 3 xul.dll!mozilla::WebGLContext::CreateAndInitGL(bool,std::vector<mozilla::WebGLContext::FailureReason,std::allocator<mozilla::WebGLContext::FailureReason> > * const) [WebGLContext.cpp:0160424142d14988f7b30595e03f156c31278a42 : 590 + 0x4a]
20:22:51 INFO - eip = 0x53b031ba esp = 0x0032d82c ebp = 0x0032d894 ebx = 0x0c530173
20:22:51 INFO - esi = 0x0032d858 edi = 0x0c53a800
20:22:51 INFO - Found by: call frame info

:jfkthame, this seems to be directly related to setting the preference, is this something you should look at or should I bug jgilbert about webgl?

Comment 2

[Joel Maher ( :jmaher ) (UTC -0800)]

Attachment: Bug 1545859 - Set gfx.font_ahem_antialias_none=True for reftests. r=jfkthame

Set gfx.font_ahem_antialias_none=True for reftests.

Comment 3

[Release mgmt bot [:sylvestre / :calixte / :marco for bugbug]]

Bugbug thinks this bug is a task, but please change it back in case of error.

Comment 4

[Jonathan Kew (:jfkthame)]

Assertion failure: false (d3dcompiler DLL loading failed.)

Doesn't seem to make any sense to me; I don't see why the gfx.font_ahem_antialias_none should have any effect on this. Maybe jgilbert will have ideas about what might cause this to fail?

Comment 5

[Jeff Gilbert [:jgilbert]]

Nothing obvious to me. Does the parent cset test green on try?

Comment 6

[Joel Maher ( :jmaher ) (UTC -0800)]

yes, I tried this a few base m-c revisions, even with artifact/no-artifact; I am really confused why this is the case. running without this change tests seem to be working just fine. I can push again with latest bits- maybe when I was testing over the weekend the machines were in a funky state?

Comment 7

[Joel Maher ( :jmaher ) (UTC -0800)]

and some fresh pushes:
https://treeherder.mozilla.org/#/jobs?repo=try&author=jmaher%40mozilla.com&fromchange=decfeb909c7692faf34161a1c5ea9f31a34da4fe&tochange=50d4bb55808c6a6c4cf1434c48ce7e67055ed0c5

will check back tonight and see if all is well.

Comment 8

[Joel Maher ( :jmaher ) (UTC -0800)]

ok, both failed and now running with --no-artifact I cannot reproduce this failure:
https://treeherder.mozilla.org/#/jobs?repo=try&author=jmaher%40mozilla.com&fromchange=0cabfec8dc912a164f5f9d6ae91e2ed1b8bcc867&tochange=3932c5a9f04efd571fe45ec49dc1fc0ff635b41c

sorry to randomize both of you on this bug over some sloppy work.

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/1c939f7211ad
Set gfx.font_ahem_antialias_none=True for reftests. r=jfkthame

Comment 10

[Oana Pop-Rus]

https://hg.mozilla.org/mozilla-central/rev/1c939f7211ad