[c++] error: address of vector element requested

[vfdff]

• test: https://gcc.godbolt.org/z/z3bnbvn14

void testVectorComponentAccess() {
  typedef float v4sf __attribute__ ((vector_size (16)));
  static v4sf q;
  float* r = &q[0]; // expected-error {{address of vector element requested}}
}

For above small case, the gcc accepts, while llvm report a error, so is it a GCC extension ?
the above case is spilted from expr-address-of.c

[llvmbot]

@llvm/issue-subscribers-clang-frontend

Author: Allen (vfdff)

* test: https://gcc.godbolt.org/z/z3bnbvn14 ``` void testVectorComponentAccess() { typedef float v4sf __attribute__ ((vector_size (16))); static v4sf q; float* r = &q[0]; // expected-error {{address of vector element requested}} } ``` For above small case, the gcc accepts, while llvm report a error, so is it a **GCC extension** ? the above case is spilted from [expr-address-of.c](https://chromium.googlesource.com/external/github.com/kripken/emscripten-fastcomp-clang/+/refs/tags/1.30.0/test/Sema/expr-address-of.c)

[shafik]

Based on the code here:

llvm-project/clang/lib/Sema/SemaExpr.cpp

Lines 15160 to 15162 in a8cb9db

	} else if (op->getObjectKind() == OK_VectorComponent) {
	// The operand cannot be an element of a vector
	AddressOfError = AO_Vector_Element;

This looks to be a purposeful choice on the clang side.

CC @efriedma-quic who added the comment here: 3a1e692

[frederick-vs-ja]

This was intended, see #2475 (from https://bugs.llvm.org/show_bug.cgi?id=2103).