Fix SkBBoxRecord to correctly handle SkRect arguments that are inverted rects.

Fix SkBBoxRecord to correctly handle SkRect arguments that are inverted rects.

BUG=http://code.google.com/p/skia/issues/detail?id=1014
TEST=gm strokerect with --tileGrid
Committed: https://code.google.com/p/skia/source/detail?r=6870

[junov1, 2012-12-17 21:41:22]

PTAL

[reed1, 2012-12-17 22:07:33]

Do we not have a problem in other sites for inverted rects? Who is the caller in
this case?

[junov1, 2012-12-17 22:14:19]

On 2012/12/17 22:07:33, reed1 wrote:
> Do we not have a problem in other sites for inverted rects? Who is the caller
in
> this case?

Several call sites need this, but not all.  Basically all the draw* methods that
receive an SkRect.

[reed1, 2012-12-17 22:19:16]

Can you create an Issue that documents what skp (I presume) triggers this, and
then note that we need a fix in all canvas subclasses? I can be the owner if you
like, but I want to capture the concrete caller/example right away.

lgtm for this change.

[junov1, 2012-12-17 22:30:41]

Other canvas subclasses are not showing an issue with this.  gm test strokerect
would fail if they did.

I think I could concoct a test that would falsely trigger SkCanvas:: quickReject
though.  I will follow-up with that.

[reed1, 2012-12-17 22:45:58]

Is the existing gm strokerect the guy who is triggering this? I'm just trying to
learn how you came across this problem.

[junov1, 2012-12-18 15:12:50]

Exactly. for the past 2 days I've been cleaning up the failures that appear
with gm --tileGrid.

This is my plan for getting SkTileGrid ready for prime-time:
Once "gm --tileGrid" is completely fixed, I will change gm to run these
tests by default.
I have only a couple of failures left to fix.  Next step is to do a similar
validation with render_pictures on our collection of skp files.
Then I will start over again using scaled playback, to validate bbox
scaling logic and AA padding.
Once all that is solved, we can turn it on in Chrome for impl-side
rendering, and hopefully nothing will explode.

On Mon, Dec 17, 2012 at 5:45 PM, <reed@google.com> wrote:

> Is the existing gm strokerect the guy who is triggering this? I'm just
> trying to learn how you came across this problem.
>
>
https://codereview.appspot.**com/6945071/<https://codereview.appspot.com/6945...
>