Review: more nan/inf checking

At Chris Kulla's suggestion, expand our "debugnan" checking to include the binding step, to catch any bad values passed into the shading system.

[ckulla, 2010-02-01 16:26:06]

http://codereview.appspot.com/196066/diff/1/3
File src/liboslexec/exec.cpp (right):

http://codereview.appspot.com/196066/diff/1/3#newcode345
src/liboslexec/exec.cpp:345: if (debugnan && check_nan (sym, NULL, 0,
m_npoints-1, badval))
Can you pass the original run flags here? Otherwise you are checking points that
might not be valid.

http://codereview.appspot.com/196066/diff/1/3#newcode384
src/liboslexec/exec.cpp:384: if (!get_renderer_userdata (m_npoints,
wants_derivatives,
It would be good to check the results here too.

[lg_imageworks.com, 2010-02-01 17:34:24]

Good call on both counts.  I will upload an updated review shortly.

	-- lg


On Feb 1, 2010, at 8:26 AM, <ckulla@gmail.com> <ckulla@gmail.com> wrote:

> 
> http://codereview.appspot.com/196066/diff/1/3
> File src/liboslexec/exec.cpp (right):
> 
> http://codereview.appspot.com/196066/diff/1/3#newcode345
> src/liboslexec/exec.cpp:345: if (debugnan && check_nan (sym, NULL, 0,
> m_npoints-1, badval))
> Can you pass the original run flags here? Otherwise you are checking
> points that might not be valid.
> 
> http://codereview.appspot.com/196066/diff/1/3#newcode384
> src/liboslexec/exec.cpp:384: if (!get_renderer_userdata (m_npoints,
> wants_derivatives,
> It would be good to check the results here too.
> 
> http://codereview.appspot.com/196066/show
> 

--
Larry Gritz
lg@imageworks.com

[larrygritz, 2010-02-01 17:53:18]

Updated with Chris Kulla's suggestions

[ckulla, 2010-02-01 18:02:50]

LGTM