If Ashmem cache fails pinCache, do not reallocate.

If Ashmem cache fails pinCache, do not reallocate.

Committed: https://code.google.com/p/skia/source/detail?r=7973

[reed1, 2013-03-04 21:05:34]

lgtm w/ question

https://codereview.chromium.org/12398021/diff/2001/src/ports/SkAshmemImageCac...
File src/ports/SkAshmemImageCache.cpp (right):

https://codereview.chromium.org/12398021/diff/2001/src/ports/SkAshmemImageCac...
src/ports/SkAshmemImageCache.cpp:53: SkASSERT(pRec->fFD != -1);
What is the return value for ashmem_pin_region that we are ignoring?

[scroggo, 2013-03-04 21:10:47]

https://codereview.chromium.org/12398021/diff/2001/src/ports/SkAshmemImageCac...
File src/ports/SkAshmemImageCache.cpp (right):

https://codereview.chromium.org/12398021/diff/2001/src/ports/SkAshmemImageCac...
src/ports/SkAshmemImageCache.cpp:53: SkASSERT(pRec->fFD != -1);
On 2013/03/04 21:05:35, reed1 wrote:
> What is the return value for ashmem_pin_region that we are ignoring?

ASHMEM_WAS_PURGED or ASHMEM_NOT_PURGED. In this case, we already discovered that
the memory was purged (which is why we call allocAndPinCache again), so we do
not care about the return value (same down below, for the newly allocated memory
in this same function).

[reed1, 2013-03-04 21:18:13]

got it. perhaps we can check it inside an assert (also documents this
assumption).

[scroggo, 2013-03-04 21:31:05]

On 2013/03/04 21:18:13, reed1 wrote:
> got it. perhaps we can check it inside an assert (also documents this
> assumption).

We don't actually know the answer. When we last called ashmem_pin_region, it
returned ASHMEM_WAS_PURGED, but then we unpinned and pinned it again. I assume
that it will then report whether it was purged again, which could be either one.

[scroggo, 2013-03-04 21:38:55]

Committed patchset #3 manually as r7973 (presubmit successful).