On Fri, 2009-08-07 at 10:12 -0700, Steven Dake wrote:
> comment on this inline
> > #if defined(HAVE_PTHREAD_SETSCHEDPARAM) &&
> > defined(HAVE_SCHED_GET_PRIORITY_MAX)
> > - if (policy != SCHED_OTHER) {
> I believe the reason this check is here, is that if SCHED_OTHER is
> specified, sched param should be null in the call to
> pthread_setschedparam. The default for a new thread is SCHED_OTHER so
> there is no need to set the parameter.
>
> Before committing, I'd like you to verify that specifying SCHED_OTHER
> works properly.
I have been look at the man page for pthread_setschedparam and the only
thing it mentions is that policy is required by RR and FIFO but doesn't
explicitly mention that policy has to be NULL for OTHER, I believe that
the policy parameter is simply ignored in case of OTHER. The
sched_setpolicy man page mentioned in the above man page seems to have
vanished from any distribution and google has no knowledge of it.
In any case I believe, since we are simply wrapping
pthread_setschedparam because we don't want to export logsys_thread_id
that is the caller responsibility to provide the right parameters to the
call and that if one day _OTHER will accept policy info, then we won't
have to change one line of code.
Anyway in my tests it works, because corosync mainexec does the right
calls by passing NULL when configuring OTHER.
Fabio
_______________________________________________
Openais mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/openais
