Add Clang Thread Safety Annotations (locks-only)

[jiridanek]

• More easily digestible subset of Add Clang Thread Safety Annotations #212

Uses https://clang.llvm.org/docs/ThreadSafetyAnalysis.html (with all the limitations that stem from it, namely, works only in clang or IDE based on clangd, and the GUARDED_BY annotation cannot be used in C structs, llvm/llvm-project#20777)

How it works in action

Example of a text-only compile failure message: https://github.com/jiridanek/skupper-router/actions/runs/4555107395/jobs/8033958647#step:25:282

Example of the clangd IDE integration:

Calling a _LH function without holding the lock

Unlocking something that is not locked

Returning without unlocking

[codecov]

Codecov Report

Merging #1036 (59e500e) into main (7c658af) will decrease coverage by 0.1%.
The diff coverage is 95.5%.

Additional details and impacted files

@@           Coverage Diff           @@
##            main   #1036     +/-   ##
=======================================
- Coverage   77.7%   77.7%   -0.1%     
=======================================
  Files        247     247             
  Lines      63991   64003     +12     
  Branches    5896    5896             
=======================================
- Hits       49775   49755     -20     
- Misses     11532   11548     +16     
- Partials    2684    2700     +16     

Flag	Coverage Δ
pysystemtests	87.7% <ø> (-0.1%)	⬇️
pyunittests	54.6% <ø> (ø)
systemtests	71.3% <95.5%> (-0.1%)	⬇️
unittests	21.9% <87.8%> (+<0.1%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Components	Coverage Δ
unittests	26.0% <87.8%> (+<0.1%)	⬆️
systemtests	78.4% <95.5%> (-0.1%)	⬇️

[jiridanek]

• Related to Runtime Debug tests to validate proper thread execution #1015

[jiridanek]

I'm reasonably confident that this will be low-noise and will not be causing annoying problems.

Judging from past history, the only way to get this in is probably to just push it to main and then see whether my confidence was justified, or it does cause some problems and needs to be reverted.

I still have the previous full PR (which contains additional annotation for running on the core thread) in the queue

• Add Clang Thread Safety Annotations #212

[ganeshmurthy]

I'm reasonably confident that this will be low-noise and will not be causing annoying problems.

Judging from past history, the only way to get this in is probably to just push it to main and then see whether my confidence was justified, or it does cause some problems and needs to be reverted.

I still have the previous full PR (which contains additional annotation for running on the core thread) in the queue

* [Add Clang Thread Safety Annotations #212](https://github.com/skupperproject/skupper-router/pull/212)

This is a big enough change that I would at least like @kgiusti to sign off on it before this lands on main

[jiridanek]

This is a big enough change that I would at least like @kgiusti to sign off on it before this lands on main

He said this about it (long time ago) #212 (review)

I really think this will be low friction in practice. That was the whole point of the "reduced, locks-only" PR, after all. And if not, I'm ready with my finger on the revert button. Either way, I really want to conclude this "initiative" in some way. At least this way it will be shown if thread-safety-annotations are beneficial or painful.