[WORKAROUND] Allow to use boost:filesystem instead of std::filesystem, [BUGFIX] check_cxx_linker_flag

[atamazov]

• Allows to use boost:filesystem instead of std::filesystem via CMake flag.
  • This is to workaround runtime issues (SIGSEGV) related to misconfigured/buggy std::filesystem on some host configurations (e.g. CentOS 8 with manually installed gcc 10).
  • Usage: cmake ... -DMIOPEN_WORKAROUND_USE_BOOST_FILESYSTEM=On ...
• By-products
  • [NFC][CMake] Improved logging in check_cxx_linker_flag
  • [NFC] Some formatting for readability

Locally tested (make check with debug configuration, make analyze etc.)

---

[Attribution] @junliume @JehandadKhan

• https://github.com/ROCm/MIOpen/labels/workaround
• https://github.com/ROCm/MIOpen/labels/non-miopen-bug
• https://github.com/ROCm/MIOpen/labels/urgency_blocker
• Proposed reviewers:
  • @BrianHarrisonAMD
  • @apwojcik
  • @junliume
• Optional reviewers:
  • @DrizztDoUrden
  • @CAHEK7
  • @shurale-nkn
  • @averinevg

[atamazov]

@BrianHarrisonAMD @apwojcik @junliume @JehandadKhan @DrizztDoUrden @CAHEK7 @shurale-nkn @averinevg Ready for reviewing, CI testing & merging.

[BrianHarrisonAMD]

Is there still more issues with std::filesystem leading to segfaults after the #3126 fix?

I was under the impression that the issue only came up when using GCC8, and not including stdc++fs, but seems like there must be some other way to have issues come up if you were hitting issues with CentOS 8 & GCC10.

[atamazov]

@BrianHarrisonAMD Yes, @junliume knows more details.

[atamazov]

https://github.com/ROCm/MIOpen/labels/TESTING_CI_PASSED except Windows, I guess.

[BrianHarrisonAMD]

LGTM

[atamazov]

@junliume @BrianHarrisonAMD @apwojcik @JehandadKhan @DrizztDoUrden @CAHEK7 @shurale-nkn @averinevg Ready for (hopefully last) reviewing round.

Contains a bugfix now https://github.com/ROCm/MIOpen/labels/bug.

I highly recommend re-running CI.

[atamazov]

@junliume It seems like the problem with CentOS8/gcc10 is not a blocker anymore? Even if so, I would recommend merging this soon because this PR contains a bugfix in check_cxx_linker_flag. Or, If we do not need boost:filesystem, then I can prepare the PR that contains only bugfixes and improvements.