[libc-commits] [libc] [libc] Proof of concept of aliasing long double math functions. (PR #132627)

[via libc-commits]

================
@@ -37,21 +37,42 @@
 
 #define LLVM_LIBC_ATTR(name) EXPAND_THEN_SECOND(LLVM_LIBC_FUNCTION_ATTR_##name)
 
-// MacOS needs to be excluded because it does not support aliasing.
-#if defined(LIBC_COPT_PUBLIC_PACKAGING) && (!defined(__APPLE__))
-#define LLVM_LIBC_FUNCTION_IMPL(type, name, arglist)                           \
+// MacOS needs to be excluded because it does not support [[gnu::aliasing]].
+#ifndef __APPLE__
+
+#if defined(LIBC_COPT_PUBLIC_PACKAGING)
+#define LLVM_LIBC_FUNCTION(type, name, arglist)                                \
   LLVM_LIBC_ATTR(name)                                                         \
   LLVM_LIBC_FUNCTION_ATTR decltype(LIBC_NAMESPACE::name)                       \
-      __##name##_impl__ __asm__(#name);                                        \
+      __##name##_impl__ asm(#name);                                            \
   decltype(LIBC_NAMESPACE::name) name [[gnu::alias(#name)]];                   \
   type __##name##_impl__ arglist
-#else
-#define LLVM_LIBC_FUNCTION_IMPL(type, name, arglist) type name arglist
-#endif
 
-// This extra layer of macro allows `name` to be a macro to rename a function.
+#define LLVM_LIBC_ALIAS(name, func)                                            \
+  decltype(LIBC_NAMESPACE::name) LIBC_NAMESPACE::name [[gnu::alias(#func)]];   \
+  asm(#name " = " #func);                                                      \
+  static_assert(true, "Require semicolon")
----------------
lntue wrote:

I found that forcing the semicolons at the end of these top level macros plays much nicer with clang-format, especially when they are inside `#if` blocks.  I started to add those to newer top level test macros, and we don't have to occasionally fight with clang-format.

https://github.com/llvm/llvm-project/pull/132627