Re: Review: extend inf/nan checking to derivatives, fix problem with arrays (issue199063)
|
cku...@...
LGTM, except for the previously existing bug
http://codereview.appspot.com/199063/diff/1/3 File src/liboslexec/exec.cpp (right): http://codereview.appspot.com/199063/diff/1/3#newcode346 src/liboslexec/exec.cpp:346: if (debugnan && check_nan (sym, m_context->m_original_runflags, There's a serious bug here I didn't notice in the first pass. We are checking the value of locals and temps that have not been initialized yet. This gives tons of false positives in actual scenes. http://codereview.appspot.com/199063/show |
|
|