tree b54750fefee2e0e0aa868d5db6f3473ad5c8df19
parent d9b62e7f6ebf8764854b7c3c7752bfce8f7adee2
author Sal <salguarnieri@google.com> 1558122040 -0700
committer GitHub <noreply@github.com> 1558122040 -0700
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsBcBAABCAAQBQJc3w44CRBK7hj4Ov3rIwAAdHIIAG1zng2099Hp/FmrA3M10X8r
 /+J9RGkf9kHZoMyfJ4YXAFuXUfMeKdh9mf/AeFhLFqIuMSSwomOKKOrePAcb33Ed
 Q693MR+rT8NhkFLjAIsephX7c+O4W1wQkU6B6dTjnMfF1vHKyh4cXFJPV38ePCSd
 7W6viIi6nkYaaGCkrGv/orEW2INYc+l/5w9Cc3eJ9j9N1YQMLHFLkiwk9ReMYFl1
 As0vGivI2GxY79MJ4NkkaGS0hFTV9Sjlmvso5LzAsGL1YsKiaTode66VcKjD/gOf
 cbtvkuFpVljFLftcg8x5mx5pG+fAasrgXXLtmYqv68CdEXPaHJfi1zecoo86Uzs=
 =7BeD
 -----END PGP SIGNATURE-----
 

fix(cxx): better support cuda extractions (#3747)

* fix(cxx): better support cuda extractions

Some cuda compilations will expand to more than a single clang invocation and
our extractor doesn't support that at the moment. Work around this by adding
--cuda-host-only to the clang args if we detect a cuda compilation. That will
force a single clang invocation.