TRAC #22676
Signed-off-by: Geoff Lang
Signed-off-by: Shannon Woods
Author: Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1996 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #22676
Signed-off-by: Geoff Lang
Signed-off-by: Shannon Woods
Author: Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1995 736b8ea6-26fd-11df-bfd4-992fa37f6226

Review URL: https://codereview.appspot.com/7798044

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1994 736b8ea6-26fd-11df-bfd4-992fa37f6226

After calling Reset in testDeviceLost to acknowledge a display mode change, D3D sometimes crashes on the next Present call on an existing swap chain. I couldn't get it to happen with nVidia. The stack looks like this with the debug runtime:

 	ntdll.dll!_ZwRaiseException@12()  + 0x12 bytes	
 	ntdll.dll!_ZwRaiseException@12()  + 0x12 bytes	
 	d3d9d.dll!__DDAssert()  + 0x1ca bytes	
 	d3d9d.dll!CSwapChain::PresentMain()  + 0xf52 bytes	
 	d3d9d.dll!CSwapChain::Present()  + 0x24 bytes	
 	libEGL.dll!egl::Surface::swapRect(int x, int y, int width, int height)  Line 458 + 0x2a bytes	C++
>	libEGL.dll!egl::Surface::swap()  Line 605	C++
 	libEGL.dll!eglSwapBuffers(void * dpy, void * surface)  Line 1063 + 0x8 bytes	C++
 	gl_wrapper.dll!gfx::EGLApiBase::eglSwapBuffersFn(void * dpy, void * surface)  Line 813 + 0x18 bytes	C++
 	gl_wrapper.dll!gfx::NativeViewGLSurfaceEGL::SwapBuffers()  Line 363 + 0x35 bytes	C++
 	gl_wrapper.dll!gfx::GLSurfaceAdapter::SwapBuffers()  Line 201 + 0x1d bytes	C++
 	content.dll!content::PassThroughImageTransportSurface::SwapBuffers()  Line 216 + 0xb bytes	C++

The error is "BitBlt or StretchBlt failed in Present".

With the release runtime either Present continually fails with E_INVALIDARG or, less commonly, crashes.

It seems that by recreating all the swap chains after reset, it recovers. I didn't call resetSwapChain because there is no need to recreate all the other surfaces.

As far as I know, this should not be necessary with WDDM.
Review URL: https://codereview.appspot.com/7567045

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1993 736b8ea6-26fd-11df-bfd4-992fa37f6226

The bump to 109 in r1984 collided with a similar version change in dx11proto branch (r1738).
Review URL: https://codereview.appspot.com/7446050

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1987 736b8ea6-26fd-11df-bfd4-992fa37f6226

When CheckDeviceStats returns S_PRESENT_MODE_CHANGED, reset display mode and retest for device lost.

Unless the display mode is reset, CheckDeviceLost can keep returning S_PRESENT_MODE_CHANGED, potentially masking a device lost on account of a GPU hang, causing ANGLE to fail to report context lost.
Review URL: https://codereview.appspot.com/7381061

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1986 736b8ea6-26fd-11df-bfd4-992fa37f6226

Review URL: https://codereview.appspot.com/7460048

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1985 736b8ea6-26fd-11df-bfd4-992fa37f6226

I will send the tests in a separate patch because I need to refactor the way compiler_tests are setup.
Review URL: https://codereview.appspot.com/7402051

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1984 736b8ea6-26fd-11df-bfd4-992fa37f6226

Review URL: https://codereview.appspot.com/7388051

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1983 736b8ea6-26fd-11df-bfd4-992fa37f6226

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1979 736b8ea6-26fd-11df-bfd4-992fa37f6226

Ensure nested ternary operator unfolding uses a new scope to avoid redefinition of the same temporary variable.

TRAC #22629
Signed-off-by: Jamie Madill
Signed-off-by: Shannon Woods
Author: Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1835 736b8ea6-26fd-11df-bfd4-992fa37f6226

It's causing trouble with Aura. Locking the screen seems to generate the display mode change device status and Aura cannot recover.

http://crbug.com/178148
Review URL: https://codereview.appspot.com/7398052

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1834 736b8ea6-26fd-11df-bfd4-992fa37f6226

Review URL: https://codereview.appspot.com/7381049

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1833 736b8ea6-26fd-11df-bfd4-992fa37f6226

1) Remove an empty file: this causes patch apply failures when updating WebKit side ANGLE.
2) Fix a 64-to-32 conversion issue.
3) Append a change to Tokenizer.l that's left out in a previous CL. (cpp is already updated).
Review URL: https://codereview.appspot.com/7378051

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1832 736b8ea6-26fd-11df-bfd4-992fa37f6226

Original patch: http://trac.webkit.org/changeset/139665
Review URL: https://codereview.appspot.com/7392044

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1831 736b8ea6-26fd-11df-bfd4-992fa37f6226

Review URL: https://codereview.appspot.com/7308107

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1830 736b8ea6-26fd-11df-bfd4-992fa37f6226

Sometimes IDirect3DDevice9Ex::CheckDeviceState returns S_PRESENT_MODE_CHANGED when the GPU has hung. It's not consistent, maybe one time out of two when I hang the GPU while running WebGL Aquarium.

This unfortunately means that changes to display mode will cause ANGLE to report context lost. Other ideas appreciated.
Review URL: https://codereview.appspot.com/7305093

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1829 736b8ea6-26fd-11df-bfd4-992fa37f6226

Review URL: https://codereview.appspot.com/7330046

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1828 736b8ea6-26fd-11df-bfd4-992fa37f6226

BUG=396
Review URL: https://codereview.appspot.com/7323055

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1827 736b8ea6-26fd-11df-bfd4-992fa37f6226

This is an incompatible API change, but one which is necessary in
order to improve correctness of the code. The API version in
ShaderLang.h is updated and, unfortunately, the define renamed to
something less ambiguous due to conflicts on some Android buildbots.
Temporary patches in Chromium and WebKit will be landed separately to
support this upgrade.

BUG=403,404,405,406,407,408,409
Review URL: https://codereview.appspot.com/7300058

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1826 736b8ea6-26fd-11df-bfd4-992fa37f6226

ProgramBinary::save writes a size_t, so ProgramBinary::load needs
to read it as the same type.

Landed for jschuh.
Original review: https://codereview.appspot.com/7304057/
Review URL: https://codereview.appspot.com/7310052

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1825 736b8ea6-26fd-11df-bfd4-992fa37f6226

Fixes new WebGL conformance tests
shader-with-illegal-default-precision.frag.html and
shader-with-illegal-default-precision.vert.html per
http://www.khronos.org/bugzilla/show_bug.cgi?id=781 . Also ran WebGL
1.0.1 conformance tests; no regressions.

BUG=none
Review URL: https://codereview.appspot.com/7302043

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1824 736b8ea6-26fd-11df-bfd4-992fa37f6226

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1823 736b8ea6-26fd-11df-bfd4-992fa37f6226

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1821 736b8ea6-26fd-11df-bfd4-992fa37f6226

TBR=alokp
Review URL: https://codereview.appspot.com/7228076

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1810 736b8ea6-26fd-11df-bfd4-992fa37f6226

With this fixed, Chrome behaves the same on WebGL conformance tests 1.0.2 with/without --enable-shader-name-hashing

BUG=
TEST=WebGL conformance tests
Review URL: https://codereview.appspot.com/7229043

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1809 736b8ea6-26fd-11df-bfd4-992fa37f6226

Trac #21120

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1807 736b8ea6-26fd-11df-bfd4-992fa37f6226

Trac #21120

    01/30/2013  dgkoch  add issue 6 and clear interactions
                        renamed to EXT_draw_buffers based on feedback
                        changed resolution of issue 3.

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1806 736b8ea6-26fd-11df-bfd4-992fa37f6226

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1805 736b8ea6-26fd-11df-bfd4-992fa37f6226

Review URL: https://codereview.appspot.com/7221045

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1804 736b8ea6-26fd-11df-bfd4-992fa37f6226

So len could equal = max_len, and at this point name[len] is out of the range.

BUG=
TEST=
Review URL: https://codereview.appspot.com/7223045

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1803 736b8ea6-26fd-11df-bfd4-992fa37f6226

BUG=none
TEST=compiled on Linux and Mac OS
Review URL: https://codereview.appspot.com/7133076

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1735 736b8ea6-26fd-11df-bfd4-992fa37f6226

this optimization level, it appears that array index clamping can be
enabled on Windows without destabilizing WebGL content.

Tested this change with array index clamping enabled with MapsGL, a
previous failing test from AlteredQualia, many Three.js demos, the
Julia Set tutorial from acko.net, the WebGL 1.0.1 conformance suite,
and other content; all worked.

BUG=none
Review URL: https://codereview.appspot.com/7197054

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1734 736b8ea6-26fd-11df-bfd4-992fa37f6226

array index clamping to use the clamp intrinsic. This works more
reliably on pure OpenGL ES devices and on Windows.

Added a mechanism in ShBuiltInResources to choose the strategy for
array index clamping.

BUG=none
TEST=various out-of-bounds array indexing tests and various WebGL content
Review URL: https://codereview.appspot.com/7194051

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1733 736b8ea6-26fd-11df-bfd4-992fa37f6226

BUG=none
TEST=compiled on Linux
Review URL: https://codereview.appspot.com/7201056

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1732 736b8ea6-26fd-11df-bfd4-992fa37f6226

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1731 736b8ea6-26fd-11df-bfd4-992fa37f6226

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1730 736b8ea6-26fd-11df-bfd4-992fa37f6226

Review URL: https://codereview.appspot.com/7193053

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1729 736b8ea6-26fd-11df-bfd4-992fa37f6226

Chrome now builds ANGLE against the Windows 8 Platform SDK.

Move the include and lib search paths to common.gypi so that downstream projects can specify the location of the Platform SDK or alternatively use the DirectX SDK instead.
Review URL: https://codereview.appspot.com/7193052

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1728 736b8ea6-26fd-11df-bfd4-992fa37f6226

This is to fix the standalone gyp build. It couldn't handle changes made in r1721. This is the latest version.
Review URL: https://codereview.appspot.com/7204049

git-svn-id: https://angleproject.googlecode.com/svn/trunk@1727 736b8ea6-26fd-11df-bfd4-992fa37f6226