diff options
author | Leon Scroggins III <scroggo@google.com> | 2017-01-27 13:16:28 -0500 |
---|---|---|
committer | Skia Commit-Bot <skia-commit-bot@chromium.org> | 2017-01-27 18:50:37 +0000 |
commit | 428865794bd76937f4ad03d791ca07cda131f83e (patch) | |
tree | 7f0829825b14bd8228082b5842db4010f12dd6f6 /bench | |
parent | c121a8849cf6d1d535e69fc3836c5720e0372a28 (diff) |
Set fOptions in SkCodec::getPixels
Subclasses sometimes try to read fOptions, but it used to not get set
for full decodes. As a result, they might be reading the Options from a
previous scanline/incremental decode. In addition to being wrong, this
is bad in the case of an fSubset pointing to a rectangle that no longer
exists.
So set fOptions in getPixels, prior to any attempts to read it by sub-
classes. Use a different workaround for the webp/incomplete bug. Set
fSubset to null prior to calling fillIncompleteImage. It can only be
non-null for webp, and in that case we do not want the fill call to
be using the subset width.
Modify the Codec_jpeg_rewind test to use an incomplete image, so that
it will also test fillIncompleteImage.
DM tests of inc0.webp and inc1.webp will verify that the incomplete
bug has not resurfaced.
BUG=skia:5772
Change-Id: If5e1e3c9a7f337183783299c0a9e58dcbbc84119
Reviewed-on: https://skia-review.googlesource.com/7682
Commit-Queue: Matt Sarett <msarett@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>
Reviewed-by: Matt Sarett <msarett@google.com>
Diffstat (limited to 'bench')
0 files changed, 0 insertions, 0 deletions