[PATCH] D80343: [mlir][gpu] Fix CMake errors in case NVPTX is not enabled. NFC.

[Mehdi AMINI via Phabricator via llvm-commits]

mehdi_amini added a comment.

I only see this now, I already reverted the offending patch I think. In the future if you have a fix for a build failure, please land it directly without code review: it is more important to unbreak everyone ASAP (either through revert or hot fix). 
You can always ping the original revision to indicate that you had to do that so that the original reviewer can take a look at it post-review.

Right now the best course of action is to squash this into your original patch and land it back (there is a comment from Richard in your original revision that should likely be addressed first).



================
Comment at: mlir/test/lib/Transforms/CMakeLists.txt:6
+endif()
+
 # Exclude tests from libMLIR.so
----------------
This is too intrusive: can we define an *empty* target `MLIRGPUtoCUDATransforms` always instead? (just not using `add_mlir_conversion_library` to do so)


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D80343/new/

https://reviews.llvm.org/D80343