fix: avoid race condition when starting bgw thread

[cnicolaescu]

Specifically on Windows, the sentry_thread_spawn macro is defined as follows:

#    define sentry__thread_spawn(ThreadId, Func, Data)                         \
        (*ThreadId = CreateThread(NULL, 0, Func, Data, 0, NULL),               \
            *ThreadId == INVALID_HANDLE_VALUE ? 1 : 0)

We can see that ThreadId is initialized with the value returned by CreateThread, but the new thread might be running before CreateThread returns. If the new thread tries to access bgw->thread_id, the value might not be initialized yet, leading to undefined behaviour / crash.

In this change a new function sentry__thread_get_current_threadid is added, which returns the current thread id in a reliable way.

[cnicolaescu]

Alternatively, the new thread could be created as suspended (passing the CREATE_SUSPENDED flag to CreateThread), then resumed in a 2nd step (using the ResumeThread API). This allows to correctly initialize bgw->thread_id before the new thread starts running.
This approach involves either:

• enhanching the sentry__thread_spawn macro
• creating the suspended thread in sentry__thread_init, then resuming it in sentry__thread_spawn

Please let me know if you'd prefer one of these 2 alternative approaches.

[Swatinem]

lgtm. If I remember correctly setting the thread name is indeed the only access of the thread_id from within the thread itself. We also only create a single transport thread, and creation should be guarded in the sentry_init lock.

[codecov-commenter]

Codecov Report

Merging #785 (730f8fd) into master (ed0a62d) will decrease coverage by 0.06%.
The diff coverage is 100.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #785      +/-   ##
==========================================
- Coverage   81.33%   81.27%   -0.07%     
==========================================
  Files          53       53              
  Lines        6929     6933       +4     
  Branches     1110     1110              
==========================================
- Hits         5636     5635       -1     
- Misses       1183     1188       +5     
  Partials      110      110              

[supervacuus]

@cnicolaescu, good catch; thx for the fix. Can you add an entry to the change log to merge this?