Re: xmutex vs xlock

[Vladimir T. <vtz...@gm...>]

On Mon, Jun 5, 2017 at 12:24 AM, Compro Prasad <com...@gm...>
wrote:

> I don't see the point in having "xmutex_t" defined as "xlock_t" and
> "xmutex_raw_t" defined as "pthread_mutex_t". I mean how is xmutex_t
> related to xmutex_raw_t ? I see it as a redundant definition which
> makes the real definition look clumsy.
>


xlock_t helps to implement THREAD-INTERRUPT (and thus CALL-WITH-TIMEOUT,
THREAD-KILL) in safe way. THREAD-INTERRUPT is implemented by sending
SIGUSR1 to the thread which then is interrupted at so called GC_SAFE
regions. Waiting with pthread_mutex_lock is such GC_SAFE region but
interrupting it may/will leave the pthread_mutex_t internals in
inconsistent state.

xlock_t helps with unblocking threads waiting on mutexes when interrupt
should be handled.