pybind11 tests are failing with obscure c++ error

[mattip]

For instance this run fails with

 31%] Building CXX object tests/CMakeFiles/pybind11_tests.dir/test_class_sh_disowning_mi.cpp.o
In file included from /usr/include/signal.h:328,
                 from /home/runner/work/binary-testing/binary-testing/tests/catch/catch.hpp:8030,
                 from /home/runner/work/binary-testing/binary-testing/tests/pure_cpp/smart_holder_poc_test.cpp:14:
/home/runner/work/binary-testing/binary-testing/tests/catch/catch.hpp:10818:58: error: call to non-‘constexpr’ function ‘long int sysconf(int)’
10818 |     static constexpr std::size_t sigStackSize = 32768 >= MINSIGSTKSZ ? 32768 : MINSIGSTKSZ;
      |                                                          ^~~~~~~~~~~
In file included from /usr/include/x86_64-linux-gnu/bits/sigstksz.h:24:
/usr/include/unistd.h:640:17: note: ‘long int sysconf(int)’ declared here
  640 | extern long int sysconf (int __name) __THROW;
      |                 ^~~~~~~
/home/runner/work/binary-testing/binary-testing/tests/catch/catch.hpp:10877:45: error: size of array ‘altStackMem’ is not an integral constant-expression
10877 |     char FatalConditionHandler::altStackMem[sigStackSize] = {};
      |                                             ^~~~~~~~~~~~
gmake[2]: *** [tests/pure_cpp/CMakeFiles/smart_holder_poc_test.dir/build.make:79: tests/pure_cpp/CMakeFiles/smart_holder_poc_test.dir/smart_holder_poc_test.cpp.o] Error 1
gmake[1]: *** [CMakeFiles/Makefile2:710: tests/pure_cpp/CMakeFiles/smart_holder_poc_test.dir/all] Error 2

I checked the workflow to make sure it is comparable to upstream, I couldn't find a significant difference. This began two weeks ago. Can I summon someone from pybind11 to take a look @HenryII maybe (sorry to bother you)?

[mattip]

I meant "can I summon someone from pybind11 to take a look at this failure, @henryiii maybe"?

[henryiii]

This is about when @rwgk merged smart_holder, which was a huge overall change. For the error, looks like MINSIGSTKSZ is dynamic if _DYNAMIC_STACK_SIZE_SOURCE is set, which happens if the GNU extensions are turned on. This seems to break catch2.

[mattip]

Thanks. How do I turn off GNU extensions?

[henryiii]

Can you enable verbose compilation so I can see the command line used to compile the file? If my guess is right, I'd expect to see -std=g++11 instead of -std=c++11.

[henryiii]

Ah, I think this was fixed in Catch2 2.13.5. We download an older version if it's missing. pybind/pybind11#5568 should fix.

[rwgk]

Thanks @henryiii! — I've never seen the error before.

[henryiii]

Should be fixed on latest master branch.

[mattip]

I triggered a build just now that should pick up the latest changes. I wonder why the upstream CI didn't see this...

[mattip]

Thanks, it is fixed. I opened a PR about the "xfail (strict)" test that is now passing pybind/pybind11#5569.

Closing.