On Tue, Mar 15, 2022 at 12:55:38PM -0600, Gavin Howard wrote:
> I am pretty sure that this is the right place and that it is a bug in
> musl.

Once more we find someone blaming the implementation for their own
mistake. From your repository:

| y_Status
| y_strucon_rwlock_wrlock(y_strucon_rwlock* l)
| {
|       y_ThreadData d = y_strucon_thread_os_get();
|
|       yc_assert(l != NULL, YC_ASSERT_NULL);
|       yc_assert(d != NULL, YC_ASSERT_NULL);
|
|       y_THREAD_TRYLOCK1(d);
|
|       y_Status s = y_strucon_rwlock_os_rdlock(l);
|
|       return s;
| }

The mistake is in this function. See what the function is called and
what subfunction it calls? No wonder multiple threads can take the lock
simultaneously.

Also, while reading I found a glaring mistake in y_THREAD_TRYUNLOCK1(),
which I shall leave as homework to the reader.

Ciao,
Markus