Aggressively assert on BITMAP object deletions

Aggressively assert on BITMAP object deletions

Add asserts on bitmap object deletions to verify that they do indeed complete
successfully.  Also, when CreateDIBSection fails, attempt to make a smaller one
to determine whether the failure is due to memory exhaution / fragmentation or
some other resource exhaution.

BUG=275046

[Vangelis Kokkevis, 2013-11-02 18:41:39]

Mike, any other ideas on things we could try to better pinpoint what sort of
resource exhaustion we're facing?

[reed1, 2013-11-04 16:49:27]

https://codereview.chromium.org/57053002/diff/50001/skia/ext/bitmap_platform_...
File skia/ext/bitmap_platform_device_win.cc (right):

https://codereview.chromium.org/57053002/diff/50001/skia/ext/bitmap_platform_...
skia/ext/bitmap_platform_device_win.cc:90:
SkASSERT(DeleteObject(bitmap_handle_));
SkASSERT becomes a no-op in release builds, so I think you want to use
SkAssertResult() which always evaluates its argument.

[Vangelis Kokkevis, 2013-11-04 18:43:27]

On 2013/11/04 16:49:27, reed1 wrote:
>
https://codereview.chromium.org/57053002/diff/50001/skia/ext/bitmap_platform_...
> File skia/ext/bitmap_platform_device_win.cc (right):
> 
>
https://codereview.chromium.org/57053002/diff/50001/skia/ext/bitmap_platform_...
> skia/ext/bitmap_platform_device_win.cc:90:
> SkASSERT(DeleteObject(bitmap_handle_));
> SkASSERT becomes a no-op in release builds, so I think you want to use
> SkAssertResult() which always evaluates its argument.

Oh good catch. Actually I need this to crash even on release builds so that we
can try to catch the errors out in the field. I converted the calls to
SK_CRASH()

[jamesr, 2013-11-13 22:31:42]

+cpu to check it over.  Do all callers to CreateHBitmap definitely crash on
failure (that seems like the only sane behavior, but worth checking).

lgtm. I think I'll try to add some more diagnostics on top of this

[jamesr, 2013-11-14 01:42:50]

I went ahead and combined this diagnostic with some other checks that cpu@ wrote
for an older crash and put the combination here:

https://codereview.chromium.org/70193011

[vangelis, 2013-11-14 01:46:20]

I like yours better! Thanks!



On Wed, Nov 13, 2013 at 5:42 PM, <jamesr@chromium.org> wrote:

> I went ahead and combined this diagnostic with some other checks that
cpu@wrote
> for an older crash and put the combination here:
>
> https://codereview.chromium.org/70193011
>
> https://codereview.chromium.org/57053002/
>

To unsubscribe from this group and stop receiving emails from it, send an email
to chromium-reviews+unsubscribe@chromium.org.

[cpu_(ooo_6.6-7.5), 2013-11-14 20:24:57]

lgtm