Fix line rendering with TileGrid and Rtree

Fix line rendering with TileGrid and Rtree

BUG=http://code.google.com/p/skia/issues/detail?id=1014
TEST=gm hairmodes with --tileGrid and --rtree
Committed: https://code.google.com/p/skia/source/detail?r=6841

[junov1, 2012-12-14 20:32:08]

PTAL

[reed1, 2012-12-14 20:43:18]

How do you want to handle hairline points? Their strokeWidth will be 0.

https://codereview.appspot.com/6947056/diff/1/src/core/SkBBoxRecord.cpp
File src/core/SkBBoxRecord.cpp (right):

https://codereview.appspot.com/6947056/diff/1/src/core/SkBBoxRecord.cpp#newco...
src/core/SkBBoxRecord.cpp:26: bbox.set(pts, count);
halfStrokeWidth = SkScalarHalf(paint.getStrokeWidth());

is simpler. Or paint.getStrokeWidth() / 2;

[junov1, 2012-12-14 21:38:15]

Patch

[reed1, 2012-12-14 21:47:24]

I know its a small change, but I find this code hard to understand why we do
what we do, and who handles what parts of the contract (e.g. who handles
antialias inflation). Hence the request for more dox.

https://codereview.appspot.com/6947056/diff/1003/src/core/SkBBoxRecord.cpp
File src/core/SkBBoxRecord.cpp (right):

https://codereview.appspot.com/6947056/diff/1003/src/core/SkBBoxRecord.cpp#ne...
src/core/SkBBoxRecord.cpp:27: // Small min width value, just to ensure hairline
point bounding boxes aren't empty.
Perhaps we can expand the comment to help explain
1. why we need something other than width/2
2. why a tiny value makes sense, rather than a min of 1.0

[junov1, 2012-12-14 22:13:28]

Patch

[reed1, 2012-12-17 12:59:57]

lgtm