[Iftop-users] iftop on openbsd

[Paul Warren]

On Fri, Aug 29, 2003 at 05:13:02PM +0100, Chris Lightfoot wrote:
> On Fri, Aug 29, 2003 at 05:07:29PM +0100, Paul Warren wrote:
> > On Fri, Aug 29, 2003 at 04:54:11PM +0100, Chris Lightfoot wrote:
> 
> > > > My guess is that it's a difference in the handling of threads (is that
> > > > vague enough?), and the stuff that makes is quit is <blame-torch
> > > > action="pass">stuff that Chris wrote</blame-torch>
>     [...]
> > At a quick guess, I'd say that pcap_loop doesn't call any thread
> > cancellation points, so it hangs on pthread_cancel.  
> 
> Ah, that's a good point. That said, I'm surprised that
> quitting works on Mac OS X if that's the problem....
> 
> Try this patch: (not absolutely certain about the second
> parameter to pcap_dispatch; try 1 if -1 doesn't work)
> 
> 
> diff -u -r1.44 iftop.c
> --- iftop.c     2003/05/20 21:14:37     1.44
> +++ iftop.c     2003/08/29 16:10:06
> @@ -528,7 +528,9 @@
>  /* packet_loop:
>   * Worker function for packet capture thread. */
>  void packet_loop(void* ptr) {
> -    pcap_loop(pd,-1,(pcap_handler)packet_handler,NULL);
> +    /* Make sure that this thread is cancellable. */
> +    while (pcap_dispatch(p, -1, (pcap_handler)packet_handler, NULL) != -1)
> +        pcap_testcancel();
>  }

ITYM pthread_testcancel, and that'll nead the thread handle.

> (I should say that I don't recall being the one to put the
> PCAP stuff in a thread. I have a horror of threads
> anyway....)

No.  I don't understand threads, but I managed to figure out just enough
to get them to start in different threads.  Obviously, I couldn't then
figure out how to make the stop again - you fixed that bit, and
therefore claim ownership of that functionality ;-)  It may be far less
broken than when you first touched it, but you were the last one to
touch it...

Paul