pthread/spinlock: Call up_testsest directly in the flat build

[xiaoxiang781216]

Summary

• libc: Move the common implementation of up_testset to libc/machine

Impact

spinlock

Testing

Pass CI

[Ouss4]

@xiaoxiang781216 a fix for the macOS failure on ESP32 configs is here: #6281 however I see other ARM related changes, seem to be build related.

[xiaoxiang781216]

@xiaoxiang781216 a fix for the macOS failure on ESP32 configs is here: #6281 however I see other ARM related changes, seem to be build related.

Done.

[Ouss4]

Change looks good.
I'm just wondering if arch_atomic.c is the best place for the up_testset function.

cc. @gustavonihei

[xiaoxiang781216]

Change looks good. I'm just wondering if arch_atomic.c is the best place for the up_testset function.

cc. @gustavonihei

up_testset is one of atomic function. Actually, it isn't good to introduce up_testset since atomic is part of C/C++ standard. It's better to use the standard atomic api, and implement them on the old toolchain by hand.

[gustavonihei]

Change looks good. I'm just wondering if arch_atomic.c is the best place for the up_testset function.
cc. @gustavonihei

up_testset is one of atomic function.

The part that makes this new location a bit weird is that arch_atomic.c is under libc, and up_testset is specific for spinlocks.

Actually, it isn't good to introduce up_testset since atomic is part of C/C++ standard. It's better to use the standard atomic api, > and implement them on the old toolchain by hand.

So another option would be to leave up_testset in its place and change it to use the C11 function. And then on arch_atomic.c we could provide the implementation for the GCC/Clang builtins for those archs that don't support.

[xiaoxiang781216]

Ok, I drop up_testset from PR, let's create a new PR to remove up_testset and call atomic api in spinlock instead.

[xiaoxiang781216]

Here is #6283

[gustavonihei]

LGTM

[Ouss4]

Here is #6283

Recursion. :)

[xiaoxiang781216]

should be #6301 :)