996266 - Do context creation attrib fallback

Status: RESOLVED FIXED

(Core :: Graphics: CanvasWebGL, defect)

Description

[Kelsey Gilbert [:jgilbert]]

Right now, we don't really fall back when context creation fails with a given set of attribs. We detect that, for example, MSAA is disabled, and force antialias:false, but we don't actually 'fall back' programmatically.

We should do this properly.

Comment 1

[Kelsey Gilbert [:jgilbert]]

Attachment: depth-fallback

:kamidphish for the WebGL, :bjacob for the GLContext. (and WebGL too, if he wants!)

Comment 2

[Benoit Jacob [:bjacob] (mostly away)]

Please share some more background:
 - is this fixing a concrete bug or rather a general sanity improvement?
 - what is the line of thinking leading to adding a new GLContextProvider::Create* function, a rather rare event (even if as a coincidence we're getting another one added on bug 994856 !)

Comment 3

[u480271]

Comment on attachment 8406463 [details] [diff] [review]
depth-fallback

Review of attachment 8406463 [details] [diff] [review]:
-----------------------------------------------------------------

::: content/canvas/src/WebGLContext.cpp
@@ +412,5 @@
> +CreateHeadlessNativeGL(bool forceEnabled, const nsCOMPtr<nsIGfxInfo>& gfxInfo, WebGLContext& webgl)
> +{
> +    if (!forceEnabled) {
> +        if (IsFeatureInBlacklist(gfxInfo, nsIGfxInfo::FEATURE_WEBGL_OPENGL)) {
> +            webgl.GenerateWarning("Refused to create native OpenGL context because of blacklisting.");

if (!forceEnable && IsFeatureInBlacklist(...

@@ +437,5 @@
> +    nsRefPtr<GLContext> gl;
> +
> +#ifdef XP_WIN
> +    if (!forceEnabled) {
> +        if (IsFeatureInBlacklist(gfxInfo, nsIGfxInfo::FEATURE_WEBGL_ANGLE)) {

Ditto

@@ +555,5 @@
> +
> +    // we should really have this behind a
> +    // |gfxPlatform::GetPlatform()->GetScreenDepth() == 16| check, but
> +    // for now it's just behind a pref for testing/evaluation.
> +    baseCaps.bpp16 = Preferences::GetBool("webgl.prefer-16bpp", false);

Should we just do it then? How much work is it?

@@ +574,5 @@
> +    // Done with baseCaps construction.
> +
> +    bool forceAllowAA = Preferences::GetBool("webgl.msaa-force", false);
> +    if (!forceAllowAA) {
> +        if (IsFeatureInBlacklist(gfxInfo, nsIGfxInfo::FEATURE_WEBGL_MSAA)) {

Ditto

@@ +615,5 @@
> +        if (!InitAndValidateGL())
> +            break;
> +
> +        return true;
> +    } while (false);

Why the do/while(false)?

Comment 4

[Kelsey Gilbert [:jgilbert]]

(In reply to Dan Glastonbury :djg :kamidphish from comment #3)
> Comment on attachment 8406463 [details] [diff] [review]
> depth-fallback
> 
> Review of attachment 8406463 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> ::: content/canvas/src/WebGLContext.cpp
> @@ +412,5 @@
> > +CreateHeadlessNativeGL(bool forceEnabled, const nsCOMPtr<nsIGfxInfo>& gfxInfo, WebGLContext& webgl)
> > +{
> > +    if (!forceEnabled) {
> > +        if (IsFeatureInBlacklist(gfxInfo, nsIGfxInfo::FEATURE_WEBGL_OPENGL)) {
> > +            webgl.GenerateWarning("Refused to create native OpenGL context because of blacklisting.");
> 
> if (!forceEnable && IsFeatureInBlacklist(...
Ok.
> 
> @@ +437,5 @@
> > +    nsRefPtr<GLContext> gl;
> > +
> > +#ifdef XP_WIN
> > +    if (!forceEnabled) {
> > +        if (IsFeatureInBlacklist(gfxInfo, nsIGfxInfo::FEATURE_WEBGL_ANGLE)) {
> 
> Ditto
Ok.
> 
> @@ +555,5 @@
> > +
> > +    // we should really have this behind a
> > +    // |gfxPlatform::GetPlatform()->GetScreenDepth() == 16| check, but
> > +    // for now it's just behind a pref for testing/evaluation.
> > +    baseCaps.bpp16 = Preferences::GetBool("webgl.prefer-16bpp", false);
> 
> Should we just do it then? How much work is it?
If we want to do it, we should do it in another bug.
> 
> @@ +574,5 @@
> > +    // Done with baseCaps construction.
> > +
> > +    bool forceAllowAA = Preferences::GetBool("webgl.msaa-force", false);
> > +    if (!forceAllowAA) {
> > +        if (IsFeatureInBlacklist(gfxInfo, nsIGfxInfo::FEATURE_WEBGL_MSAA)) {
> 
> Ditto
Ok.
> 
> @@ +615,5 @@
> > +        if (!InitAndValidateGL())
> > +            break;
> > +
> > +        return true;
> > +    } while (false);
> 
> Why the do/while(false)?
It makes cleanup easy, instead of having to cleanup in each if, or using real gotos.

Comment 5

[u480271]

(In reply to Jeff Gilbert [:jgilbert] from comment #4)
> > @@ +615,5 @@
> > > +        if (!InitAndValidateGL())
> > > +            break;
> > > +
> > > +        return true;
> > > +    } while (false);
> > 
> > Why the do/while(false)?
> It makes cleanup easy, instead of having to cleanup in each if, or using
> real gotos.

Oh right. I was having a stupid moment.

Comment 6

[Benoit Jacob [:bjacob] (mostly away)]

Jeff, did you see my questions in comment 2?

Comment 7

[Kelsey Gilbert [:jgilbert]]

(In reply to Benoit Jacob [:bjacob] from comment #2)
> Please share some more background:
>  - is this fixing a concrete bug or rather a general sanity improvement?
>  - what is the line of thinking leading to adding a new
> GLContextProvider::Create* function, a rather rare event (even if as a
> coincidence we're getting another one added on bug 994856 !)

1. Both. Right now, if creating a stencil-only context fails, we'll completely fail to create the context, instead of falling back to trying without stencil.

2. This is one step closer to moving all the Offscreen stuff to WebGLContext, which removes a ton of mostly webgl-only code from GLContext. It also reflects how we want to create things. Really, we either want a context for a widget, or a headless context, which we'll do all our own stuff for.

Comment 8

[Kelsey Gilbert [:jgilbert]]

https://tbpl.mozilla.org/?tree=Try&rev=41cad8ab68d1

Comment 9

[Kelsey Gilbert [:jgilbert]]

https://tbpl.mozilla.org/?tree=Try&rev=28007c053750

Comment 10

[Kelsey Gilbert [:jgilbert]]

https://tbpl.mozilla.org/?tree=Try&rev=be1093c11b05

Comment 11

[Kelsey Gilbert [:jgilbert]]

I filed: https://code.google.com/p/angleproject/issues/detail?id=718

As seen on:
https://tbpl.mozilla.org/?tree=Try&rev=1391ea2979ab

Comment 12

[Kelsey Gilbert [:jgilbert]]

https://tbpl.mozilla.org/?tree=Try&rev=bf50995bc6aa

Comment 13

[Kelsey Gilbert [:jgilbert]]

Attachment: patch 1: Fallback from context creation failure.

Unfortunately, I didn't keep the old patch free of changes, so we have to review it again. :(

Comment 14

[Kelsey Gilbert [:jgilbert]]

Attachment: patch 2: More fixes

Comment 15

[Kelsey Gilbert [:jgilbert]]

Attachment: patch 3: Fix ANGLE

I'm actually not sure how we want to handle ANGLE bugs at this point.
I'll figure this out.

Comment 17

[u480271]

Comment on attachment 8468985 [details] [diff] [review]
patch 1: Fallback from context creation failure.

Review of attachment 8468985 [details] [diff] [review]:
-----------------------------------------------------------------

::: dom/canvas/WebGLContext.cpp
@@ +512,5 @@
> +                       const nsCOMPtr<nsIGfxInfo>& gfxInfo,
> +                       WebGLContext* webgl)
> +{
> +    if (!forceEnabled) {
> +        if (IsFeatureInBlacklist(gfxInfo, nsIGfxInfo::FEATURE_WEBGL_OPENGL)) {

if (!forceEnabled && IsFeatureInBlacklist(...)) {
}

@@ +540,5 @@
> +    nsRefPtr<GLContext> gl;
> +
> +#ifdef XP_WIN
> +    if (!forceEnabled) {
> +        if (IsFeatureInBlacklist(gfxInfo, nsIGfxInfo::FEATURE_WEBGL_ANGLE)) {

Ditto

@@ +671,5 @@
> +    // Done with baseCaps construction.
> +
> +    bool forceAllowAA = Preferences::GetBool("webgl.msaa-force", false);
> +    if (!forceAllowAA) {
> +        if (IsFeatureInBlacklist(gfxInfo, nsIGfxInfo::FEATURE_WEBGL_MSAA)) {

if (!forceAllowAA && IsFeatureInBlacklist(...)) {
}

::: gfx/gl/GLContext.cpp
@@ +1090,5 @@
>          raw_fGetIntegerv(LOCAL_GL_SCISSOR_BOX, mScissorRect);
>          raw_fGetIntegerv(LOCAL_GL_MAX_TEXTURE_SIZE, &mMaxTextureSize);
>          raw_fGetIntegerv(LOCAL_GL_MAX_CUBE_MAP_TEXTURE_SIZE, &mMaxCubeMapTextureSize);
>          raw_fGetIntegerv(LOCAL_GL_MAX_RENDERBUFFER_SIZE, &mMaxRenderbufferSize);
>  

Drop the blank line?

Comment 18

[u480271]

Comment on attachment 8468986 [details] [diff] [review]
patch 2: More fixes

Review of attachment 8468986 [details] [diff] [review]:
-----------------------------------------------------------------

::: gfx/angle/src/libGLESv2/renderer/d3d9/RenderTarget9.cpp
@@ +14,5 @@
>  #include "libGLESv2/renderer/d3d9/renderer9_utils.h"
>  #include "libGLESv2/renderer/d3d9/formatutils9.h"
>  #include "libGLESv2/main.h"
>  
> +#include <cstdio>

o_O

::: gfx/gl/GLBlitHelper.cpp
@@ +43,5 @@
>                GLenum aType, const gfx::IntSize& aSize, bool linear)
>  {
>      GLuint tex = 0;
>      aGL->fGenTextures(1, &tex);
> +        ScopedBindTexture autoTex(aGL, tex);

Indent

::: gfx/gl/GLContext.cpp
@@ +1116,5 @@
>          raw_fGetIntegerv(LOCAL_GL_MAX_VIEWPORT_DIMS, mMaxViewportDims);
>  
> +        printf_stderr("mMaxTextureSize: %d\n", mMaxTextureSize);
> +        printf_stderr("mMaxRenderbufferSize: %d\n", mMaxRenderbufferSize);
> +        printf_stderr("mMaxViewportDims: %dx%d\n", mMaxViewportDims[0],

Should this be disabled? (I'm not sure what the rules are about printing at runtime)

@@ +1160,5 @@
>          mMaxSamples = 0;
>          if (IsSupported(GLFeature::framebuffer_multisample)) {
>              fGetIntegerv(LOCAL_GL_MAX_SAMPLES, (GLint*)&mMaxSamples);
>          }
> +        printf_stderr("mMaxSamples: %d\n", mMaxSamples);

Ditto

::: gfx/gl/GLContext.h
@@ +522,4 @@
>  // -----------------------------------------------------------------------------
>  // Error handling
>  public:
> +    static const char* GLErrorToString(GLenum aError) {

I can never tell whether we put { on the same line for functions or on the next.

Comment 19

[u480271]

Comment on attachment 8468988 [details] [diff] [review]
patch 3: Fix ANGLE

Review of attachment 8468988 [details] [diff] [review]:
-----------------------------------------------------------------

Do you need to contribute this to Vlad's mozilla patches to ANGLE on github?

Comment 20

[Kelsey Gilbert [:jgilbert]]

(In reply to Dan Glastonbury :djg :kamidphish from comment #17)
> Comment on attachment 8468985 [details] [diff] [review]
> patch 1: Fallback from context creation failure.
> 
> Review of attachment 8468985 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> ::: dom/canvas/WebGLContext.cpp
> @@ +512,5 @@
> > +                       const nsCOMPtr<nsIGfxInfo>& gfxInfo,
> > +                       WebGLContext* webgl)
> > +{
> > +    if (!forceEnabled) {
> > +        if (IsFeatureInBlacklist(gfxInfo, nsIGfxInfo::FEATURE_WEBGL_OPENGL)) {
> 
> if (!forceEnabled && IsFeatureInBlacklist(...)) {
> }
Fixed.
> 
> @@ +540,5 @@
> > +    nsRefPtr<GLContext> gl;
> > +
> > +#ifdef XP_WIN
> > +    if (!forceEnabled) {
> > +        if (IsFeatureInBlacklist(gfxInfo, nsIGfxInfo::FEATURE_WEBGL_ANGLE)) {
> 
> Ditto
Fixed.
> 
> @@ +671,5 @@
> > +    // Done with baseCaps construction.
> > +
> > +    bool forceAllowAA = Preferences::GetBool("webgl.msaa-force", false);
> > +    if (!forceAllowAA) {
> > +        if (IsFeatureInBlacklist(gfxInfo, nsIGfxInfo::FEATURE_WEBGL_MSAA)) {
> 
> if (!forceAllowAA && IsFeatureInBlacklist(...)) {
> }
Fixed.
> 
> ::: gfx/gl/GLContext.cpp
> @@ +1090,5 @@
> >          raw_fGetIntegerv(LOCAL_GL_SCISSOR_BOX, mScissorRect);
> >          raw_fGetIntegerv(LOCAL_GL_MAX_TEXTURE_SIZE, &mMaxTextureSize);
> >          raw_fGetIntegerv(LOCAL_GL_MAX_CUBE_MAP_TEXTURE_SIZE, &mMaxCubeMapTextureSize);
> >          raw_fGetIntegerv(LOCAL_GL_MAX_RENDERBUFFER_SIZE, &mMaxRenderbufferSize);
> >  
> 
> Drop the blank line?
Fixed.

Comment 21

[Kelsey Gilbert [:jgilbert]]

(In reply to Dan Glastonbury :djg :kamidphish from comment #18)
> Comment on attachment 8468986 [details] [diff] [review]
> patch 2: More fixes
> 
> Review of attachment 8468986 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> ::: gfx/angle/src/libGLESv2/renderer/d3d9/RenderTarget9.cpp
> @@ +14,5 @@
> >  #include "libGLESv2/renderer/d3d9/renderer9_utils.h"
> >  #include "libGLESv2/renderer/d3d9/formatutils9.h"
> >  #include "libGLESv2/main.h"
> >  
> > +#include <cstdio>
> 
> o_O
Hahaha, sorry. I was calling std::fprintf(stderr,...) here, as I needed to debug via logging on Try. ><
> 
> ::: gfx/gl/GLBlitHelper.cpp
> @@ +43,5 @@
> >                GLenum aType, const gfx::IntSize& aSize, bool linear)
> >  {
> >      GLuint tex = 0;
> >      aGL->fGenTextures(1, &tex);
> > +        ScopedBindTexture autoTex(aGL, tex);
> 
> Indent
Fixed.
> 
> ::: gfx/gl/GLContext.cpp
> @@ +1116,5 @@
> >          raw_fGetIntegerv(LOCAL_GL_MAX_VIEWPORT_DIMS, mMaxViewportDims);
> >  
> > +        printf_stderr("mMaxTextureSize: %d\n", mMaxTextureSize);
> > +        printf_stderr("mMaxRenderbufferSize: %d\n", mMaxRenderbufferSize);
> > +        printf_stderr("mMaxViewportDims: %dx%d\n", mMaxViewportDims[0],
> 
> Should this be disabled? (I'm not sure what the rules are about printing at
> runtime)
Removed.
> 
> @@ +1160,5 @@
> >          mMaxSamples = 0;
> >          if (IsSupported(GLFeature::framebuffer_multisample)) {
> >              fGetIntegerv(LOCAL_GL_MAX_SAMPLES, (GLint*)&mMaxSamples);
> >          }
> > +        printf_stderr("mMaxSamples: %d\n", mMaxSamples);
> 
> Ditto
Removed.
> 
> ::: gfx/gl/GLContext.h
> @@ +522,4 @@
> >  // -----------------------------------------------------------------------------
> >  // Error handling
> >  public:
> > +    static const char* GLErrorToString(GLenum aError) {
> 
> I can never tell whether we put { on the same line for functions or on the
> next.
We put it on the same line if the function decl is indented at all, and fits on one line.
Therefore it's on the same line in inline member function definitions, but CPP file definitions of member functions have it on the next line.

Comment 22

[Kelsey Gilbert [:jgilbert]]

Attachment: patch 1: Fallback from context creation failure.

r=kamidphish

Comment 23

[Kelsey Gilbert [:jgilbert]]

Attachment: patch 2: More fixes

Comment 24

[Kelsey Gilbert [:jgilbert]]

(In reply to Dan Glastonbury :djg :kamidphish from comment #19)
> Comment on attachment 8468988 [details] [diff] [review]
> patch 3: Fix ANGLE
> 
> Review of attachment 8468988 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> Do you need to contribute this to Vlad's mozilla patches to ANGLE on github?

I think so.

Comment 25

[Kelsey Gilbert [:jgilbert]]

  https://tbpl.mozilla.org/?tree=Try&rev=6cf2bbd51902

Comment 26

[u480271]

Comment on attachment 8469513 [details] [diff] [review]
patch 2: More fixes

Review of attachment 8469513 [details] [diff] [review]:
-----------------------------------------------------------------

::: gfx/gl/GLContext.cpp
@@ +281,1 @@
>      mHasRobustness(false),

Try is complaining about this:

/builds/slave/try-lx-d-000000000000000000000/build/gfx/gl/GLContext.h:618:10: error: 'mozilla::gl::GLContext::mIsInLocalErrorCheck' will be initialized after [-Werror=reorder]
/builds/slave/try-lx-d-000000000000000000000/build/gfx/gl/GLContext.h:520:10: error: 'bool mozilla::gl::GLContext::mHasRobustness' [-Werror=reorder]
/builds/slave/try-lx-d-000000000000000000000/build/gfx/gl/GLContext.cpp:268:1: error: when initialized here [-Werror=reorder]

Comment 27

[Kelsey Gilbert [:jgilbert]]

(In reply to Dan Glastonbury :djg :kamidphish from comment #26)
> Comment on attachment 8469513 [details] [diff] [review]
> patch 2: More fixes
> 
> Review of attachment 8469513 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> ::: gfx/gl/GLContext.cpp
> @@ +281,1 @@
> >      mHasRobustness(false),
> 
> Try is complaining about this:
> 
> /builds/slave/try-lx-d-000000000000000000000/build/gfx/gl/GLContext.h:618:10:
> error: 'mozilla::gl::GLContext::mIsInLocalErrorCheck' will be initialized
> after [-Werror=reorder]
> /builds/slave/try-lx-d-000000000000000000000/build/gfx/gl/GLContext.h:520:10:
> error: 'bool mozilla::gl::GLContext::mHasRobustness' [-Werror=reorder]
> /builds/slave/try-lx-d-000000000000000000000/build/gfx/gl/GLContext.cpp:268:
> 1: error: when initialized here [-Werror=reorder]

Fixed in:
https://tbpl.mozilla.org/?tree=Try&rev=ae72b8551f1f

Comment 28

[Kelsey Gilbert [:jgilbert]]

Comment on attachment 8469513 [details] [diff] [review]
patch 2: More fixes

Is this r+ with the fully-passing Try run?

Comment 29

[u480271]

(In reply to Jeff Gilbert [:jgilbert] from comment #28)
> Comment on attachment 8469513 [details] [diff] [review]
> patch 2: More fixes
> 
> Is this r+ with the fully-passing Try run?

Yes. Don't you just need to switch order of 

    #ifdef DEBUG
        mIsInLocalErrorCheck(false),
    #endif
    mHasRobustness(false),

to what it was?

Comment 30

[Kelsey Gilbert [:jgilbert]]

(In reply to Dan Glastonbury :djg :kamidphish from comment #29)
> (In reply to Jeff Gilbert [:jgilbert] from comment #28)
> > Comment on attachment 8469513 [details] [diff] [review]
> > patch 2: More fixes
> > 
> > Is this r+ with the fully-passing Try run?
> 
> Yes. Don't you just need to switch order of 
> 
>     #ifdef DEBUG
>         mIsInLocalErrorCheck(false),
>     #endif
>     mHasRobustness(false),
> 
> to what it was?

Yes. My local patch has this. It's in the TBPL link. I'll find it and upload it.

Comment 31

[Kelsey Gilbert [:jgilbert]]

Attachment: patch 2: Support checking for errors at the GLContext level

Comment 32

[Kelsey Gilbert [:jgilbert]]

Attachment: patch 3: E_INVALIDARG sometimes is an OOM/refusal to allocate

Comment 33

[Kelsey Gilbert [:jgilbert]]

Attachment: patch 0: Kill WS

Comment 34

[Kelsey Gilbert [:jgilbert]]

Using E_INVALIDARG instead of D3DERR_INVALIDCALL gives us green Win8 Debug:
https://tbpl.mozilla.org/?tree=Try&rev=4b8d623acc1a

Comment 35

[u480271]

Comment on attachment 8475664 [details] [diff] [review]
patch 2: Support checking for errors at the GLContext level

Review of attachment 8475664 [details] [diff] [review]:
-----------------------------------------------------------------

::: gfx/gl/SharedSurfaceANGLE.cpp
@@ +151,5 @@
>                       EGLConfig config,
>                       const gfx::IntSize& size)
>  {
> +    auto width = size.width;
> +    auto height = size.height;

Why?

Comment 36

[u480271]

Comment on attachment 8475665 [details] [diff] [review]
patch 3: E_INVALIDARG sometimes is an OOM/refusal to allocate

Review of attachment 8475665 [details] [diff] [review]:
-----------------------------------------------------------------

I wonder if there's a cap in DX9 for the maximum memory size a texture can have?

Comment 37

[u480271]

Comment on attachment 8475667 [details] [diff] [review]
patch 0: Kill WS

Review of attachment 8475667 [details] [diff] [review]:
-----------------------------------------------------------------

::: modules/libpref/init/all.js
@@ +2106,5 @@
>  // but provides more time for other operations when the browser is
>  // heavily loaded.
>  pref("layout.frame_rate.precise", false);
>  
> +// pref to control whether layout warnings that are hit quite often are enabled

To the god of WS we say "not today!"

Comment 38

[Kelsey Gilbert [:jgilbert]]

(In reply to Dan Glastonbury :djg :kamidphish from comment #35)
> Comment on attachment 8475664 [details] [diff] [review]
> patch 2: Support checking for errors at the GLContext level
> 
> Review of attachment 8475664 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> ::: gfx/gl/SharedSurfaceANGLE.cpp
> @@ +151,5 @@
> >                       EGLConfig config,
> >                       const gfx::IntSize& size)
> >  {
> > +    auto width = size.width;
> > +    auto height = size.height;
> 
> Why?

I think I only had this pulled out to allow printfs. It's not really hurting, even if it's a little more verbose. I'll remove it.

Comment 39

[Kelsey Gilbert [:jgilbert]]

Angle patch merged here:
https://github.com/mozilla/angle/pull/6

Comment 40

[Kelsey Gilbert [:jgilbert]]

 https://hg.mozilla.org/integration/mozilla-inbound/rev/fdd2bb040eb6
 https://hg.mozilla.org/integration/mozilla-inbound/rev/6bcda9f223b3
 https://hg.mozilla.org/integration/mozilla-inbound/rev/c97ad94c5220
 https://hg.mozilla.org/integration/mozilla-inbound/rev/2984228b8fb1

Comment 41

[Carsten Book [:Tomcat]]

sorry had to backout the above 3 csets (except the Kill WS change) for bustage on B2g like https://tbpl.mozilla.org/php/getParsedLog.php?id=46841680&tree=Mozilla-Inbound

Comment 42

[Kelsey Gilbert [:jgilbert]]

*shakes fist at b2g's GCC*

Comment 43

[Kelsey Gilbert [:jgilbert]]

 https://tbpl.mozilla.org/?tree=Try&rev=8d4cc011be81

Comment 44

[Kelsey Gilbert [:jgilbert]]

It builds now:
https://hg.mozilla.org/integration/mozilla-inbound/rev/1b09de02394a
https://hg.mozilla.org/integration/mozilla-inbound/rev/fe6593bd7068
https://hg.mozilla.org/integration/mozilla-inbound/rev/cbc1d3ef1d97
https://hg.mozilla.org/integration/mozilla-inbound/rev/626bcfb3724b

Comment 45

[Kelsey Gilbert [:jgilbert]]

Attachment: patch 2.1: Fix compiler bustage for B2G GCC.

Comment 46

[Kelsey Gilbert [:jgilbert]]

Comment on attachment 8475667 [details] [diff] [review]
patch 0: Kill WS

This stuck, and so is checked in.

Comment 47

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/fdd2bb040eb6

Comment 48

[Kelsey Gilbert [:jgilbert]]

My bad, I should have made this keep-open.

Comment 49

[Kelsey Gilbert [:jgilbert]]

Oops, almost forgot these:
https://hg.mozilla.org/integration/mozilla-inbound/rev/1b09de02394a
https://hg.mozilla.org/integration/mozilla-inbound/rev/fe6593bd7068
https://hg.mozilla.org/integration/mozilla-inbound/rev/cbc1d3ef1d97
https://hg.mozilla.org/integration/mozilla-inbound/rev/626bcfb3724b

Comment 50

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/1b09de02394a
https://hg.mozilla.org/mozilla-central/rev/fe6593bd7068
https://hg.mozilla.org/mozilla-central/rev/cbc1d3ef1d97
https://hg.mozilla.org/mozilla-central/rev/626bcfb3724b