llvm dev - Dec 2016 - failing bootstrap: C++11 or greater is required but the compiler does not support c++11

started with
http://lab.llvm.org:8011/builders/sanitizer-x86_64-linux-fuzzer/builds/1380
(Fri Dec 9 18:42:04 2016)
The previous one (Fri Dec 9 17:41:56 2016) was fine.
There was just one revision between these runs (mine), but reverting it
does not help.
Maybe an update in Ubuntu 14.04 broke this?

On Sat, Dec 10, 2016 at 3:35 PM, Eric Fiselier <eric at efcs.ca> wrote:
> I can't reproduce right away. I wonder if the check for the -std=c++11
> flag is failing due to the earlier error about failing to find libatomic.
>
> When these bots start failing?
>
> /Eric
>
> On Fri, Dec 9, 2016 at 10:18 PM, Kostya Serebryany via llvm-dev <
> llvm-dev at lists.llvm.org> wrote:
>
>> several hours ago one of the bots that does a bootstrap started to fail
>> (http://lab.llvm.org:8011/builders/sanitizer-x86_64-linux-fuzzer)
>>
>> CMake Error at projects/libcxx/CMakeLists.txt:365 (message):
>>   C++11 or greater is required but the compiler does not support c++11
>>
>>
>> I can reproduce it locally with
>>
>> cmake -GNinja -DCMAKE_CXX_COMPILER=clang++ ~/llvm
>>
>>
>> Goes away if I remove -GNinja. (!!!)
>> Does anyone else see this?
>>
>>
>> _______________________________________________
>> LLVM Developers mailing list
>> llvm-dev at lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20161210/1ed04ebb/attachment.html>

So it seems that libatomic went missing between build #1379 and #1380, so I
don't think this is related to the -std=c++11 failure. Instead it seems
likely that the compile test for -std=c++11 is failing due to
mis-configuring -latomic.

Can you confirm the bot has libatomic installed?
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20161211/6eb14df6/attachment-0001.html>

On Dec 11, 2016, at 3:33 AM, Eric Fiselier via llvm-dev <llvm-dev at
lists.llvm.org> wrote:
> So it seems that libatomic went missing between build #1379 and #1380, so I
don't think this is related to the -std=c++11 failure. Instead it seems
likely that the compile test for -std=c++11 is failing due to mis-configuring
-latomic.
> 
> Can you confirm the bot has libatomic installed?
You shouldn't need it on x86-64. The issue there is that cmake is reporting,
incorrectly:
-- Performing Test LIBCXX_HAVE_CXX_ATOMICS_WITHOUT_LIB
-- Performing Test LIBCXX_HAVE_CXX_ATOMICS_WITHOUT_LIB - Failed

despite that previously in the same run, the same test configuring llvm itself
worked:
-- Performing Test HAVE_CXX_ATOMICS_WITHOUT_LIB
-- Performing Test HAVE_CXX_ATOMICS_WITHOUT_LIB - Success

I suspect that the flags change has somehow caused *all* compiler invocations to
fail, and that results in both the erroneously-failed atomic check and the
erroneously-failed std=c++11 check.