[PATCH] D39481: [CodeGen] fix const-ness of builtin equivalents of <math.h> and <complex.h> functions that might set errno

[Eli Friedman via Phabricator via cfe-commits]

efriedma added a comment.

> But we don't check the const attribute in CGBuiltin.cpp before converting it to an intrinsic

In practice, fma() implementations don't set errno (at least, glibc and msvcrt don't).  And people would be really annoyed if we forced them to use fast-math flags to generate an fma instruction.  So this is probably okay?


https://reviews.llvm.org/D39481