[clang] Do not infer lifetimebound for functions with void return type

[usx95]

Fixes: #126231
Also found in : microsoft/STL#5271

[llvmbot]

@llvm/pr-subscribers-clang

Author: Utkarsh Saxena (usx95)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/131997.diff

1 Files Affected:

• (modified) clang/lib/Sema/SemaAttr.cpp (+1-1)

diff --git a/clang/lib/Sema/SemaAttr.cpp b/clang/lib/Sema/SemaAttr.cpp
index defdda17c32de..a6e827a421134 100644
--- a/clang/lib/Sema/SemaAttr.cpp
+++ b/clang/lib/Sema/SemaAttr.cpp
@@ -217,7 +217,7 @@ void Sema::inferGslOwnerPointerAttribute(CXXRecordDecl *Record) {
 }
 
 void Sema::inferLifetimeBoundAttribute(FunctionDecl *FD) {
-  if (FD->getNumParams() == 0)
+  if (FD->getNumParams() == 0 || FD->getReturnType()->isVoidType())
     return;
 
   if (unsigned BuiltinID = FD->getBuiltinID()) {

[hokein]

It would be nice to have have a test for it. I think it should not be too hard to add one.

[ilya-biryukov]

LGTM!
I second the suggestion to add a test. Also see the suggestion to turn this into a warning in the first place: #126231 (comment)

I am not sure why lifetimebound needs to be an error while other attributes like const produce a warning and can be ignored.

[ilya-biryukov]

There seem to be some errors missing the attribute tests now (see buildkite failures)

[ilya-biryukov]

@usx95 I struggled to come up with a reproducer and couldn't understand why at first.
I think we infer the attribute after we check for this error. This is obviously something we need to fix, but you can get a reproducer like this:

https://gcc.godbolt.org/z/Wf3eqMabE

// RUN: %clang_cc1 -std=c++20 -verify %s

namespace std {
template <class T>
constexpr const T& as_const(T&) noexcept;

// We need two declarations to see the error for some reason *shrug*
template <class T> void as_const(const T&&) noexcept = delete;
template <class T> void as_const(const T&&) noexcept;
}

namespace GH126231 {

void test() {
  int a = 1;
  std::as_const(a);
}
}

[cor3ntin]

/cherry-pick 65ee281

[llvmbot]

/pull-request #133997