Unifex with fno exceptions

[BartolomeyKant]

Add ability to build with -fno-exceptions.

In general this only fixes use try/catch to UNIFEX_TRY/UNIFEX_CATCH and wrap some code in #if !UNIFEX_NO_EXCEPTIONS ... #endif.

Also this adds run CI testing with -fno-exceptions flag.

This PR related to #671 but still doesn't give the answer to main question.

[MaxMotovilov]

Would you mind running clang-format -i on all modified files and then resubmitting? Spurious formatting changes make the diff harder to review.

[BartolomeyKant]

Would you mind running clang-format -i on all modified files and then resubmitting? Spurious formatting changes make the diff harder to review.

I did! I mean my IDE configured to automatically format on save.

But it makes a lot of changes. I discarded most of them and tried to format manually only that I really intend to change.
It looks like some of them accidentally was added to commit. I will fix it too.

I'm using clang-format --version

clang-format version 21.1.6

[MaxMotovilov]

There are some apparently spurious changes in the code that need an explanation; also it is not at all clear how sync_wait() is expected to support set_error() completion.

[MaxMotovilov]

This is a pretty major change - what was the reason behind it?

[BartolomeyKant]

Because there is mismatched declaration with detail/unifex_fwd.hpp:76.

I checked other declarations like connect and start and all of it go with inline const.

Btw. I suppose it throws error only on my machine.
I use clang 21.1.6, and gcc 15.2.1.

[MaxMotovilov]

Any chance you could demonstrate this in a godbolt? Simply including both headers does not currently trigger it w/21.1 and 15.2 respectively although your updated versions could be more recent.

If that's not possible, could you paste the error from your run?

[BartolomeyKant]

I was wrong! It's neither clang nor gcc. It was msvc with /std:c++17.
The error message looks like:

  [1/229] Building CXX object source\CMakeFiles\unifex.dir\async_manual_reset_event.cpp.obj
  FAILED: source/CMakeFiles/unifex.dir/async_manual_reset_event.cpp.obj 
  C:\PROGRA~1\MICROS~2\2022\COMMUN~1\VC\Tools\MSVC\1444~1.352\bin\Hostx64\x64\cl.exe  /nologo /TP  -IC:\projects\libunifex\include -IC:\projects\libunifex\out\build\x64-Debug\include /DWIN32 /D_WINDOWS /W3 /GR /EHsc /MDd /Zi /Ob0 /Od /RTC1 -std:c++17 /W3 /WX /bigobj /showIncludes /Fosource\CMakeFiles\unifex.dir\async_manual_reset_event.cpp.obj /Fdsource\CMakeFiles\unifex.dir\unifex.pdb /FS -c C:\projects\libunifex\source\async_manual_reset_event.cpp
C:\projects\libunifex\include\unifex\scheduler_concepts.hpp(192): error C2370: 'unifex::schedule': redefinition; different storage class
  C:\projects\libunifex\include\unifex/detail/unifex_fwd.hpp(76): note: see declaration of 'unifex::schedule'
  [2/229] Building CXX object source\CMakeFiles\unifex.dir\async_auto_reset_event.cpp.obj
  FAILED: source/CMakeFiles/unifex.dir/async_auto_reset_event.cpp.obj 
  C:\PROGRA~1\MICROS~2\2022\COMMUN~1\VC\Tools\MSVC\1444~1.352\bin\Hostx64\x64\cl.exe  /nologo /TP  -IC:\projects\libunifex\include -IC:\projects\libunifex\out\build\x64-Debug\include /DWIN32 /D_WINDOWS /W3 /GR /EHsc /MDd /Zi /Ob0 /Od /RTC1 -std:c++17 /W3 /WX /bigobj /showIncludes /Fosource\CMakeFiles\unifex.dir\async_auto_reset_event.cpp.obj /Fdsource\CMakeFiles\unifex.dir\unifex.pdb /FS -c C:\projects\libunifex\source\async_auto_reset_event.cpp
C:\projects\libunifex\include\unifex\scheduler_concepts.hpp(192): error C2370: 'unifex::schedule': redefinition; different storage class
  C:\projects\libunifex\include\unifex/detail/unifex_fwd.hpp(76): note: see declaration of 'unifex::schedule'
  [3/229] Building CXX object googletest-build\googletest\CMakeFiles\gtest.dir\src\gtest-all.cc.obj
C:\projects\libunifex\out\build\x64-Debug\cl : Command line warning D9025: overriding '/W4' with '/W3'
  ninja: build stopped: subcommand failed.

I can revert this change and fix it when I figured out what's wrong with building on msvc /std:c++17

Any chance you could demonstrate this in a godbolt?

For some reason I could not make an example with msvc on godbolt. https://godbolt.org/z/sb48aYav6

[BartolomeyKant]

BTW. With this fix msvc still fails with a ton of different errors.

msvc_build_failed.log

[MaxMotovilov]

See my new inline comments - also I think macOS clang makes an objection? https://github.com/facebookexperimental/libunifex/actions/runs/21514257277/job/62090827242?pr=673

[BartolomeyKant]

I think macOS clang makes an objection? https://github.com/facebookexperimental/libunifex/actions/runs/21514257277/job/62090827242?pr=673

I did not touch this test on this PR. And it also runs successfully on my machine and my repository.

101/101 Test  #40: test-async_mutex_v2_test ................***Timeout 1500.18 sec
Running main() from googletest-src/googletest/src/gtest_main.cc
[==========] Running 6 tests from 2 test suites.
[----------] Global test environment set-up.
[----------] 1 test from async_mutex_v2
[ RUN      ] async_mutex_v2.multiple_threads

I would like to rerun this test. And maybe someone find how to fix this bug in the future.