On Nov 1, 2021, at 10:20 PM, Brad Smith <brad@comstyle.com> wrote:
>
>
> On 10/28/2021 8:42 AM, Jeremie Courreges-Anglas wrote:
>> On Thu, Oct 28 2021, Jeremie Courreges-Anglas <jca@wxcvbn.org> wrote:
>>> On Thu, Oct 28 2021, Christian Weisgerber <naddy@mips.inka.de> wrote:
>>>> I ran an amd64 bulk build with base clang updated to LLVM 13.
>>>>
>>>> There is substantial fallout. Java appears to be broken and there
>>>> are numerous failures caused by clang 13 issuing new warnings and
>>>> ports using -Werror.
>>>>
>>>> Failure logs:
>>>> http://build-failures.rhaalovely.net/amd64-clang/2021-10-27/
>>>>
>>>> The final error may be misleading. For instance, devel/libdsm fails
>>>> with a linker error "undefined reference to pthread_create", but
>>>> the root cause is -Werror breaking a configure check.
>>>>
>>>> Below there's a list of affected ports. Several hundred more weren't
>>>> built because they depend on something that failed.
>>>>
>>>> audio/cmus
>>> jsg@ pointed out that this new clang defines __STDC_NO_ATOMICS__
>>> which is most likely wrong. This breaks at least cmus:
>>>
>>> http://build-failures.rhaalovely.net/amd64-clang/2021-10-27/audio/cmus.log
>> Brad, are you sure that this change is 100% correct?
>> https://github.com/llvm/llvm-project/commit/d8cd7806310c51af912a647a6ca46de62ff13214
>>
>> Should we drop the __STDC_NO_ATOMICS__ define?
>
> The atomics part was my misunderstanding and is wrong. I have reverted it upstream and will
> have it merged into 13.
Thanks for fixing that. Can you explain why __STDC_NO_THREADS__ was retained? I'm
assuming that's disabling std::thread and related thread support which also seems wrong.
-Kurt
No comments:
Post a Comment