On Jan 13, 2012, at 12:40 PM, Nick Mathewson wrote:
>> if (ev == NULL) {
>> /* if no time-based events are active wait for I/O */
>> fprintf(stderr, "NO TIME-BASED EVENTS ACTIVE - tvp %d\n", (tv ==
>> NULL) ? -1 : (int)tv->tv_sec);
>> goto out;
>> }
>>
>
> That's not Libevent's code! Libevent, even in 2.0.13-stable, has:
>
> if (ev == NULL) {
> /* if no time-based events are active wait for I/O */
> *tv_p = NULL;
> goto out;
> }
>
Hah! I see what happened. I removed it because it was marked as an OMPI change:
- /**** OMPI CHANGE ****/
- /* set a timeval to avoid blocking select on Windows */
-#ifndef WIN32
- *tv_p = NULL;
-#else
- struct timeval now = {1, 0};
- **tv_p = now;
-#endif
- /**** END OMPI CHANGE ****/
In fact, the OMPI change is only the correction for Windows, which doesn't like
a NULL timeval pointer. I'll restore that and make the comment clearer, then
test and get back to you.
Did you absorb all the Windows changes we sent you (believe this was one, but
there are others)? Or do you want them again?
> Are you absolutely sure you were right when you said:
>
> "Also fwiw: the OMPI changes are confined to configuration/Makefile
> areas - we actually don't fiddle with the libevent code itself other
> than a couple of places where we test for stdbool.h before including
> it."
>
> Because it seems like somewhere along the line, you lost the "*tv_p = NULL;"
>
> yrs,
> --
> Nick
> ***********************************************************************
> To unsubscribe, send an e-mail to majord...@freehaven.net with
> unsubscribe libevent-users in the body.
***********************************************************************
To unsubscribe, send an e-mail to majord...@freehaven.net with
unsubscribe libevent-users in the body.
