[PATCH] D140247: Disable Univariate3D_Invert to workaround module build failure

Steven Wu via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue Dec 20 10:49:05 PST 2022


steven_wu added inline comments.


================
Comment at: llvm/unittests/Support/LinearPolyBaseTest.cpp:171
 
+#ifndef __APPLE__
+// FIXME: workaround for module build failures:
----------------
sdesmalen wrote:
> I'm not sure that I understand either the problem or the solution :)
> 
> Is there a bug in Clang that this doesn't compile correctly when using modules? Or is there perhaps something wrong with the implementation of `operator-` (e.g. it not being correct/compliant yet GCC/Clang always accepted it).
> 
> Is this failure somehow specific to Darwin?
I think Darwin bots is the only one that runs module build from a bootstrapped compiler. I don't know this affect other platform, most likely yes. Unfortunately we don't have a flag to check if module is enabled (?). This is not a bug fix, just to get the bots green again.

This is a bug in clang module I think. @vsapsai can comment on it more but some changes in clang frontend triggered the failure. Unfortunately, the bots was broken for a long time so we can't revert cleanly the offending commits when we find it. Temporarily disable this test on Darwin bots so we can caught regression like this earlier. 


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D140247/new/

https://reviews.llvm.org/D140247



More information about the llvm-commits mailing list