[libcxx-commits] [PATCH] D119173: [libc++] Remove _LIBCPP_ABI_UNSTABLE

Wed Feb 9 05:41:30 PST 2022

ldionne added a comment.

In D119173#3306641 <https://reviews.llvm.org/D119173#3306641>, @paulkirth wrote:

> Hi,
>
> We're seeing breakage in Fuchsia's Clang canary builders for Linux (aarch64 & X86_64) that seem related to this patch, (though it may be related to D119036 <https://reviews.llvm.org/D119036> ? I can't tell).
>
> Sorry for the late notification. This was hidden by a breakage in an different LLVM component, so we missed this for a bit.
> [...]

Thanks for the heads up. It was indeed caused by D119036 <https://reviews.llvm.org/D119036>, and should be fixed by eceb40183cbb9442c7f42e7a4b8324abb50c32ab <https://reviews.llvm.org/rGeceb40183cbb9442c7f42e7a4b8324abb50c32ab>.

Can you folks move the Fuchsia bots to our pre-commit CI infrastructure?

In D119173#3305963 <https://reviews.llvm.org/D119173#3305963>, @ronlieb wrote:

> Hi 
> Looks like hip buildbot breaks from this 
> https://lab.llvm.org/buildbot/#/builders/165/builds/15363

Indeed, you need to nuke your `CMakeCache.txt`. Generally speaking, a build bot that reuses an existing cache is brittle and you should be re-performing the CMake configure step each time. If you are trying to increase caching, IMO you're better off with something like `ccache`.

================
Comment at: libcxx/CMakeLists.txt:190-192
+if (NOT LIBCXX_ABI_NAMESPACE MATCHES "__.*")
+  message(FATAL_ERROR "LIBCXX_ABI_NAMESPACE must be a reserved identifier.")
+endif()
----------------
lebedev.ri wrote:
> This requires build dir purge and cmake rerun,
> because `LIBCXX_ABI_NAMESPACE` would be `""` (empty) before,
> and you don't force-recover it to `"__${LIBCXX_ABI_VERSION}"`.
Technically you only need to clear your `CMakeCache.txt`.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D119173/new/

https://reviews.llvm.org/D119173