[PATCH] D102820: [Clang] Check for returns_nonnull when deciding to add allocation null checks

[Di Mo via Phabricator via cfe-commits]

modimo added a comment.

In D102820#2772184 <https://reviews.llvm.org/D102820#2772184>, @bruno wrote:

> Sounds reasonable to me! Can you double check whether this attribute gets correctly serialized/deserialized in face of `CXXNewExpr`? An example of how to test that would be in `clang/test/PCH/cxx-method.cpp`.

Piggybacking on that test case:

Inputs/cxx-method.h:

  typedef __typeof__(sizeof(0)) size_t;
  
  void *operator new(size_t size)
  {
    return ::operator new(size);
  }

cxx-method.cpp:

  int * foo()
  {
      return new int;
  }

Testing

  ~/llvm-project/clang/test/PCH# ~/llvm-project/build-rel/bin/clang++ -cc1 -x c++  -emit-pch Inputs/cxx-method.h -o test.pch
  ~/llvm-project/clang/test/PCH# ~/llvm-project/build-rel/bin/clang++ -cc1 -x c++ cxx-method.cpp -include-pch test.pch -emit-llvm
  ~/llvm-project/clang/test/PCH# grep _Znwm cxx-method.ll
  define dso_local nonnull i8* @_Znwm(i64 %size) #0 {
    %call = call noalias nonnull i8* @_Znwm(i64 %0) #2
    %call = call noalias nonnull i8* @_Znwm(i64 4) #3

Assuming I'm answering the correct question that the returns_nonnull is preserved through a PCH, the answer is yes.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D102820/new/

https://reviews.llvm.org/D102820