Use double precision when iterating in SkDashPathEffect::filterPath()

Use double precision when iterating in SkDashPathEffect::filterPath()

Extremely large path_length/dash_length ratios may cause us to loop
indefinitely otherwise.

R=reed@google.com

BUG=

Committed: https://code.google.com/p/skia/source/detail?r=6773

[reed1, 2012-12-12 21:21:01]

how slow is the test (in debug build)?

https://codereview.appspot.com/6926051/diff/1/src/effects/SkDashPathEffect.cpp
File src/effects/SkDashPathEffect.cpp (right):

https://codereview.appspot.com/6926051/diff/1/src/effects/SkDashPathEffect.cp...
src/effects/SkDashPathEffect.cpp:188: 
Please add a comment here what and why you're doing, and possibly a reference to
the bug and/or unittest that shows the need for using doubles.

https://codereview.appspot.com/6926051/diff/1/tests/DrawPathTest.cpp
File tests/DrawPathTest.cpp (right):

https://codereview.appspot.com/6926051/diff/1/tests/DrawPathTest.cpp#newcode229
tests/DrawPathTest.cpp:229: SkScalar intervals[] = { 0.2, 0.2 };
We will get warnings unless you use 0.2f

[fmalita, 2012-12-12 21:41:01]

On 2012/12/12 21:21:01, reed1 wrote:
> how slow is the test (in debug build)?

About 5s, vs. 0.25s previously. I can try to shave off some iterations, but
there's not much margin left for triggering the bug.

But after the dash limiting fix
(http://code.google.com/p/chromium/issues/detail?id=165432) lands, this should
get much faster: the test will be effectively short-circuited - and that is one
of the reasons I wanted to land this first and verify that it fixes the problem
on all platforms.

[reed1, 2012-12-12 21:58:41]

lgtm w/ comment suggestion

Can the unittest itself reference the Issue# ?

https://codereview.appspot.com/6926051/diff/6001/src/effects/SkDashPathEffect...
File src/effects/SkDashPathEffect.cpp (right):

https://codereview.appspot.com/6926051/diff/6001/src/effects/SkDashPathEffect...
src/effects/SkDashPathEffect.cpp:191: // (for extreme path_length/dash_length
ratios).
// See test_infinite_dash() unittest.

[fmalita, 2012-12-12 22:10:47]

On 2012/12/12 21:58:41, reed1 wrote:
> lgtm w/ comment suggestion
> 
> Can the unittest itself reference the Issue# ?
> 
>
https://codereview.appspot.com/6926051/diff/6001/src/effects/SkDashPathEffect...
> File src/effects/SkDashPathEffect.cpp (right):
> 
>
https://codereview.appspot.com/6926051/diff/6001/src/effects/SkDashPathEffect...
> src/effects/SkDashPathEffect.cpp:191: // (for extreme path_length/dash_length
> ratios).
> // See test_infinite_dash() unittest.

Thanks, I'll update it.