move failable work outside of constructors for SkGlyphCache and SkScalerContextGDI. If we fail, try…

move failable work outside of constructors for SkGlyphCache and SkScalerContextGDI. If we fail, try to purge everything and try again, in case the failure was from memory/os-resource exhaustion (e.g. out of HDC handles in windows)

BUG=

Committed: https://code.google.com/p/skia/source/detail?r=9930

[bungeman-skia, 2013-07-08 19:36:49]

It seems like what is wanted is for every time one tries to obtain an handle and
it fails, call a HandleGC() method which tries to garbage collect handles (from
caches in general) and then tries again. I'm not sure how valid purging the
glyph cache indirectly from a scaler context constructor is though.

https://codereview.chromium.org/18132009/diff/1/src/ports/SkFontHost_win.cpp
File src/ports/SkFontHost_win.cpp (right):

https://codereview.chromium.org/18132009/diff/1/src/ports/SkFontHost_win.cpp#...
src/ports/SkFontHost_win.cpp:596: fFont = CreateFontIndirect(&lf);
This line also creates a handle, we should also be checking that (we just don't
currently have an assert for it, so we won't be seeing it).

[reed1, 2013-07-09 12:55:37]

On 2013/07/08 19:36:49, bungeman1 wrote:
> It seems like what is wanted is for every time one tries to obtain an handle
and
> it fails, call a HandleGC() method which tries to garbage collect handles
(from
> caches in general) and then tries again. I'm not sure how valid purging the
> glyph cache indirectly from a scaler context constructor is though.

Agreed, I didn't want to introduce reentrancy complications, hence the current
pattern.

> 
> https://codereview.chromium.org/18132009/diff/1/src/ports/SkFontHost_win.cpp
> File src/ports/SkFontHost_win.cpp (right):
> 
>
https://codereview.chromium.org/18132009/diff/1/src/ports/SkFontHost_win.cpp#...
> src/ports/SkFontHost_win.cpp:596: fFont = CreateFontIndirect(&lf);
> This line also creates a handle, we should also be checking that (we just
don't
> currently have an assert for it, so we won't be seeing it).

I will add an assert.

[reed1, 2013-07-09 14:12:14]

removed win specific aspects

[reed1, 2013-07-09 15:29:36]

Committed patchset #2 manually as r9930 (presubmit successful).