[musl] Always cancelasync as a boolean. NFC

[sbc100]

This is how all the other code in musl already treats it.

[kripken]

Hmm, this is what I see when I grep?

system/lib/libc/musl/src/thread/pthread_create.c:	self->cancelasync = 0;
system/lib/libc/musl/src/thread/__timedwait.c:	    self->cancelasync) {
system/lib/libc/musl/src/thread/__timedwait.c:				// In the case of cancelasync or PTHREAD_CANCEL_ENABLE we can just call
system/lib/libc/musl/src/thread/pthread_setcanceltype.c:	if (old) *old = self->cancelasync;
system/lib/libc/musl/src/thread/pthread_setcanceltype.c:	self->cancelasync = new;
system/lib/libc/musl/src/thread/pthread_cancel.c:	// by cancelling cooperatively when self->cancelasync is set.
system/lib/libc/musl/src/thread/pthread_cancel.c:	if (self->canceldisable == PTHREAD_CANCEL_ENABLE || self->cancelasync)
system/lib/libc/musl/src/thread/pthread_cancel.c:	if (self->cancelasync) {
system/lib/libc/musl/src/thread/pthread_cancel.c:	// See comment above about cancelasync under emscripten.
system/lib/libc/musl/src/thread/pthread_cancel.c:	if (self->cancel && (self->cancelasync || !self->canceldisable))
system/lib/libc/musl/src/thread/pthread_cancel.c:		if (t->canceldisable == PTHREAD_CANCEL_ENABLE && t->cancelasync)
system/lib/libc/musl/src/internal/pthread_impl.h:	volatile unsigned char canceldisable, cancelasync;
system/lib/libc/musl/src/time/timer_create.c:	self->cancelasync = 0;

Can't pthread_setcanceltype write a non-boolean value there?

[kripken]

Or do you mean all reads treat it as a boolean, not altogether?

[sbc100]

Can't pthread_setcanceltype write a non-boolean value there?

Yes but that would be meaningless/undefined.

The two possible values are:

#define PTHREAD_CANCEL_DEFERRED 0                                                    
#define PTHREAD_CANCEL_ASYNCHRONOUS 1    

if you pass 2 I'm not sure what it would mean, but musl happen to treat that as non-zero/PTHREAD_CANCEL_ASYNCHRONOUS everywhere.

[sbc100]

Or do you mean all reads treat it as a boolean, not altogether?

All the checks are withing the musl codebase are basically boolean checks.

Sure you could pass an invalid value to pthread_setcanceltype but that would be undefined behaviour (this change actually makes that undefined behaviour consistent between emscripten-written code and musl-internal code).

[kripken]

Fair enough. I think it is cleaner to use the constant names, but I guess this API is not going to change, and following musl's conventions is reasonable.