On Wed, Feb 3, 2010 at 2:30 AM, Steven Dake <[email protected]> wrote:
> > >
> > > Thank you very much. I added a test program and some comments on
> > > bugzilla. Also I tested my two proposals since I need some fix right
> > > now. It worked OK. Only thing I needed to export
> > > HA_cluster_type=openais. I added this to /etc/init.d/corosync before
> > > start(). I could change the pacemaker source to use openais by default
> > > if HA_cluster_type is not set, then I don't need to export. But it's
> > > OK. Until the fix is available, I will use my workaround.
> > >
> > > hj
> >
> > Would you please test the attached fix. It should provide a workaround.
> >
>
> hj
>
> Another option for workaround attached which I prefer you test instead.
> glibc's tzset call has code in it to check if it has been called, and if
> so, to exit. This results in tzset's getenv being executed before any
> threads are spawned, vs in a thread context inside a syslog call stack.
>
> Thanks
>
> That's very good idea, but it won't work. The __tzset() calls
tzset_internal(1, 1). It calls with always = 1 which forces to call getenv()
all the time unfortunately!
========== code from glibc =============
tzset_internal(always, explicit) {
----
if (is_initialized && !always)
return;
is_initialized = 1;
tz = getenv("TZ");
----
}
_______________________________________________
Openais mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/openais
