CUDA: Now pass correct FLAGS when device link cuda executables.

Previously we had a two issues when building cuda executables
that required separable compilation. The first was that we didn't
propagate FLAGS causing any -arch / -gencode flags to be dropped, and
secondly generators such as ninja would use the CXX language flags
instead of CUDA when the executable was mixed language.

1 files changed, 4 insertions, 0 deletions

diff --git a/Source/cmLinkLineComputer.h b/Source/cmLinkLineComputer.h
index bb13717..57a70bc 100644
--- a/Source/cmLinkLineComputer.h
+++ b/Source/cmLinkLineComputer.h
@@ -11,6 +11,7 @@
 #include "cmStateDirectory.h"
 
 class cmComputeLinkInformation;
+class cmGeneratorTarget;
 class cmOutputConverter;
 
 class cmLinkLineComputer
@@ -36,6 +37,9 @@ public:
   virtual std::string ComputeLinkLibraries(cmComputeLinkInformation& cli,
                                            std::string const& stdLibString);
 
+  virtual std::string GetLinkerLanguage(cmGeneratorTarget* target,
+                                        std::string const& config);
+
 protected:
   std::string ComputeLinkLibs(cmComputeLinkInformation& cli);
   std::string ComputeRPath(cmComputeLinkInformation& cli);