Refactor HAVE_PTHREAD and _POSIX_THREADS

[gojimmypi]

Description

This started as a corner case of cross-compiling testwolfcrypt, in my case for the Xtensa ESP32-S3 Linux kernel.

I didn't have a pthread.h and noticed that several other places (but not all?) are gated with HAVE_PTHREAD.

This PR refactors the HAVE_PTHREAD to not only the test.h that I encountered, but also all other places that the conditional #if defined(_POSIX_THREADS) was used.

Upon merging there will be only one check of _POSIX_THREADS in settings.h and all other code shall be using HAVE_PTHREAD to determine if POSIX threading code should be gated in.

Note that MINGW is not POSIX and is part of the global logic that will disable HAVE_PTHREAD.

The key part of this update is this compiler directive code segment added to wolfssl/wolfcrypt/setthings.h to determine if we have POSIX Threads enabled (aka HAVE_PTHREAD is defined):

/* AFTER user_settings.h is loaded,
** determine if POSIX multi-threaded: HAVE_PTHREAD  */
#if defined(SINGLE_THREADED) || defined(__MINGW32__)
    /* Never HAVE_PTHREAD in single thread, or non-POSIX mode.
    ** Reminder: MING32 is win32 threads, not POSIX threads */
    #undef HAVE_PTHREAD
#else
    #ifdef _POSIX_THREADS
        /* HAVE_PTHREAD == POSIX threads capable and enabled. */
        #undef HAVE_PTHREAD
        #define HAVE_PTHREAD 1
    #else
        /* Not manually disabled, but POSIX threads not found. */
        #undef HAVE_PTHREAD
    #endif
#endif

Fixes zd# n/a/

Testing

How did you test?

Ran the test app in cross-compiled app as well as Linux & ESP-IDF for ESP32.

Default config, threading enabled:

./configure &&  make && ./wolfcrypt/test/testwolfcrypt

Forced single-thread. This was observed to have failed during PRB with prior changes:

 ./configure --enable-singlethreaded --disable-fastmath --enable-smallstack --enable-smallstackcache  --disable-shared  && make && ./wolfcrypt/test/testwolfcrypt 

Checklist

• added tests
• updated/added doxygen
• updated appropriate READMEs
• Updated manual and documentation

[gojimmypi]

After seeing the PRB failures, it seems there was some logic missing from a few of the examples. In particular, when the threading capability exists but is explicitly turned off with --enable-singlethreaded to force single-threaded mode.

I sprinkled some && !defined(SINGLE_THREADED) logic in places where if defined(_POSIX_THREADS) is used. It appears this logic was already in use in other places.

There may be an argument to combine all that logic into a single global definition.

[dgarske]

This looks good to me. Now that this is in settings.h I think it best to change the _POSIX_THREADS elsewhere to use HAVE_PTHREAD. I also think you should add && !defined(SINGLE_THREADED) and remove it from the refactor you did.

[gojimmypi]

cool. I've applied your suggested changes and changed the PR description & title to reflect the global nature of this change

[gojimmypi]

@dgarske - After completing the desired changes for HAVE_PTHREAD and _POSIX_THREADS, I noticed there's a potentially undesired WOLFSSL_PTHREADS that should also be refactored? See the codebase for references.

If you agree, changes to (or more specifically: removal of) WOLFSSL_PTHREADS would probably be in a separate PR & I'd be happy to work on that right away.

[dgarske]

Very nice cleanup!

[gojimmypi]

See also #6545