Per https://reviews.llvm.org/D136005#3866692 the introduction of the MLIRSparseTensorEnums target in D136002 caused breakage on some versions of cmake. This differential aims to fix those errors.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
mlir/lib/Dialect/SparseTensor/IR/CMakeLists.txt | ||
---|---|---|
21 | please put a a bit more comment on why we need a phony target |
mlir/lib/Dialect/SparseTensor/IR/CMakeLists.txt | ||
---|---|---|
21 | The conditional and phony target are just more copypasta from add_mlir_library. I'm not sure if they're actually needed/helpful here, but added them just in case. (Afaict, the conditional in add_mlir_library is to prevent add_mlir_library_install causing additional errors whenever the underlying call to add_llvm_library fails. But since I'm using add_library directly rather than going through add_llvm_library, I don't know if that's actually a concern here or not.) The real/main change of this cl is just splitting the target_sources call out from the add_library call. But I'll add some more commentary to explain things. |
I can't seem to find which version the phabricator buildbot is using, but locally I'm using 3.24.2. However, the LLVM project is supposed to work for any version >=3.13.4 (https://llvm.org/docs/GettingStarted.html#software), so it's definitely something I need to fix on our end.
I'm not sure which version changed things to allow setting the CXX_STANDARD property for INTERFACE libraries, so I'll just remove it for now and leave a comment about why. Can you try the new version of this differential?
Tightening the version numbers in the comment about splitting add_library and target_sources
This fixed our bots. You might have gotten some failure mails from us but that was a networking issue, the builds are are green now. Thanks!
please put a a bit more comment on why we need a phony target