[PATCH] D101806: [TargetLowering] Only inspect attributes in the arguments for ArgListEntry

Arthur Eubanks via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Thu May 27 11:28:55 PDT 2021


aeubanks added a comment.

In D101806#2785168 <https://reviews.llvm.org/D101806#2785168>, @uweigand wrote:

> I looks like this patch, when committed as 1c7f32334d4becc725b9025fd32291a0e5729acd <https://reviews.llvm.org/rG1c7f32334d4becc725b9025fd32291a0e5729acd>, caused two additional build bot failures on s390x:
> https://lab.llvm.org/buildbot/#/builders/94/builds/3829
>  FAIL: libFuzzer:: msan.test
>  FAIL: libFuzzer:: swap-cmp.test
> This was unfortunately hidden by other failures at the time, so I didn't notice earlier.
>
> I see that https://reviews.llvm.org/D102667 was an attempt to fix problems with this patch for msan, but it addressed only the zeroext attribute.  On s390x we also often require the signext attribute, in particular for a plain "int" function argument or return value.  I have not looked into this in detail, but I see in msan_interface_internal.h quite a number of routines using "int", so it wouldn't surprise me if this could explain the regression ...

if the msan runtime functions already required signext to work, then this change wouldn't have done anything since the msan runtime functions created in MemorySanitizer.cpp were never marked with signext. See `MaybeWarningFn` and `MaybeStoreOriginFn` in MemorySanitizer.cpp


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D101806/new/

https://reviews.llvm.org/D101806



More information about the llvm-commits mailing list