Added optional "ambiguous" outgoing argument to XRay queries so that...

Added optional "ambiguous" outgoing argument to XRay queries so that
calling code may choose different y-coordinates for better robustness.
Tested and verified manually inside O3D.

BUG=none
TEST=none


Committed: https://code.google.com/p/skia/source/detail?r=586

[kbr1, 2010-07-07 18:24:22]

Please review. Thanks.

[Stephen White, 2010-07-07 18:32:21]

I'm not sure enough about this code to change its behaviour in the absence of
the ambiguity parameter.

http://codereview.appspot.com/1695051/diff/1/3
File src/core/SkGeometry.cpp (right):

http://codereview.appspot.com/1695051/diff/1/3#newcode64
src/core/SkGeometry.cpp:64: }
Hmm.  This looks like it will return different values than it did before, even
without the ambiguity parameter.  You're also using equality of floating point
values, which can be problematic.  Are you sure this is correct?

[kbr1, 2010-07-07 20:02:08]

http://codereview.appspot.com/1695051/diff/1/3
File src/core/SkGeometry.cpp (right):

http://codereview.appspot.com/1695051/diff/1/3#newcode64
src/core/SkGeometry.cpp:64: }
On 2010/07/07 18:32:21, Stephen White wrote:
> Hmm.  This looks like it will return different values than it did before, even
> without the ambiguity parameter.  You're also using equality of floating point
> values, which can be problematic.  Are you sure this is correct?

It should return the same result as before. The only difference is that the new
code explicitly checks the incoming Y against the second endpoint of the line,
in order to see whether the result is ambiguous. This case was previously
handled by the full line segment evaluation.

Using equality of floating point values is sufficient for the cases I've tested
this code with. We can change the implementation without changing the interface
in the future if it turns out this is inadequate.

[Stephen White, 2010-07-07 20:42:43]

LGTM.

http://codereview.appspot.com/1695051/diff/1/3
File src/core/SkGeometry.cpp (right):

http://codereview.appspot.com/1695051/diff/1/3#newcode64
src/core/SkGeometry.cpp:64: }
On 2010/07/07 20:02:08, kbr1 wrote:
> On 2010/07/07 18:32:21, Stephen White wrote:
> > Hmm.  This looks like it will return different values than it did before,
even
> > without the ambiguity parameter.  You're also using equality of floating
point
> > values, which can be problematic.  Are you sure this is correct?
> 
> It should return the same result as before. The only difference is that the
new
> code explicitly checks the incoming Y against the second endpoint of the line,
> in order to see whether the result is ambiguous. This case was previously
> handled by the full line segment evaluation.

OK, I see that now.

> Using equality of floating point values is sufficient for the cases I've
tested
> this code with. We can change the implementation without changing the
interface
> in the future if it turns out this is inadequate.
> 

OK, LGTM.