make GrContext::contextLost work for context destruction

[junov1, 2011-06-08 21:05:23]

PTAL.  This is for release blocker
http://code.google.com/p/chromium/issues/detail?id=85309

[bsalomon, 2011-06-08 21:14:02]

On 2011/06/08 21:05:23, junov1 wrote:
> PTAL.  This is for release blocker
> http://code.google.com/p/chromium/issues/detail?id=85309

Minor nit: braces on one line "if".

If it works I'm ok with it as a stop-gap (assuming it fixes the blocker). Long
term we'll need to flesh this out so that other calls on GrContext exit early
when there is no underlying 3D context. We should have a well defined
missing-3DCtx state for the GrContext.

[junov1, 2011-06-08 21:23:01]

Fix is incomplete. I just found a different test case that shows that shader
cache is not purged by contextLost.  More coming...

[reed1, 2011-06-08 21:29:23]

http://codereview.appspot.com/4584048/diff/1/gpu/include/GrContext.h
File gpu/include/GrContext.h (right):

http://codereview.appspot.com/4584048/diff/1/gpu/include/GrContext.h#newcode71
gpu/include/GrContext.h:71: void contextLost(bool haveNewContext = true);
Would prefer something more clear (bools are indescript at the call-site, and
default args are more so, unless its obj -vs- null).

How about:

enum ContextChange {
    kLost_ContextChange,
    kChanged_ContextChange
};
void contextChanged(...);

or other similar names...

http://codereview.appspot.com/4584048/diff/1/gpu/src/GrContext.cpp
File gpu/src/GrContext.cpp (right):

http://codereview.appspot.com/4584048/diff/1/gpu/src/GrContext.cpp#newcode98
gpu/src/GrContext.cpp:98: this->setupDrawBuffer();
always use { } for if-statements.

[bsalomon, 2011-06-08 22:23:26]

LGTM