[PATCH] D150144: cmake: add missing dependency on Attributes.inc

NAKAMURA Takumi via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue May 9 14:17:59 PDT 2023


chapuni added inline comments.


================
Comment at: llvm/utils/TableGen/CMakeLists.txt:90-91
+
+  DEPENDS
+  attributes_gen
   )
----------------
jroelofs wrote:
> chapuni wrote:
> > bulbazord wrote:
> > > bulbazord wrote:
> > > > Is this right? llvm-tblgen depends on a file generated by llvm-tblgen?
> > > Ok, I looked into this further. All the headers that get generated (everything in `llvm/include/llvm/`) are generated by `llvm-min-tblgen`, which is defined above. It **does** make sense for `llvm-tblgen` to depend on `attributes_gen`. This is not a concern anymore.
> > @jroelofs Could you explain its motivation or background, please?
> > 
> > AFAIK this is redundant in normal build. I have confirmed ToT `llvm-tblgen` doesn't depend on `Core` or `intrinsics_gen`.
> > 
> > I've noticed it might be required with `LLVM_ENABLE_MODULES`.  Not certain yet though.
> The change in this file was motivated by the modules build having broken.
> 
> > @bulbazord:
> > llvm/util/Tablegen/AsmMatcherEmitter.cpp includes llvm/util/Tablegen/CodeGenRegisters.h, which includes some MC header... which includes an IR header, which depends on this .inc file being generated.
> 
> The change in the other file was somewhat speculative, based on one of the AnalysisTests sources having included something that intrinsics_gen produces.
Thanks for your explanation. I knew it when I pruned kludges in D146352 (modmap) but I forgot restoring in D148769.
This is better than restoring modmap since it is not a big burden.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D150144/new/

https://reviews.llvm.org/D150144



More information about the llvm-commits mailing list