[libcxx-commits] [PATCH] D64378: [libcxx] Use C11 thread API on Fuchsia

[Petr Hosek via Phabricator via libcxx-commits]

phosek added inline comments.


================
Comment at: libcxx/include/__config:1068
 #  if defined(__FreeBSD__) || \
-      defined(__Fuchsia__) || \
       defined(__wasi__) || \
----------------
EricWF wrote:
> Switching the implementation used by Fuchsia is ABI breaking. Are we 100% certain that all of our Fuchsia users don't expect ABI stability?
Yes, that's fundamental part of Fuchsia's packaging story.


================
Comment at: libcxx/include/__config:1107
 #if (defined(_LIBCPP_HAS_THREAD_API_PTHREAD) && defined(__GLIBC__)) \
+  || defined(_LIBCPP_HAS_THREAD_API_FUCHSIA) \
   || defined(_LIBCPP_HAS_THREAD_API_WIN32)
----------------
EricWF wrote:
> Can you point towards the C11 source for `mtx_destroy` and `cnd_destroy`?
I've added guard for Fuchsia since this is implementation of details our libc.


================
Comment at: libcxx/include/__config:1115
 // as well.
-#if defined(_LIBCPP_HAS_THREAD_API_WIN32)
+#if defined(_LIBCPP_HAS_THREAD_API_FUCHSIA) || \
+    defined(_LIBCPP_HAS_THREAD_API_WIN32)
----------------
EricWF wrote:
> After switching to the `_LIBCPP_HAS_THREAD_API_C11` this check should include `_LIBCPP_HAS_MUSL_LIBC`, because AFAIK, the trivial destruction property is an implementation detail of MUSL.
I've instead added a check for `defined(__Fuchsia__)`, Fuchsia's libc started as a fork of musl but it has sufficiently diverged at this point that I don't think it's safe to consider it being musl-compatible.


Repository:
  rCXX libc++

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D64378/new/

https://reviews.llvm.org/D64378