Re: Running ping, need help

[Shachar Shemesh]

Oron Peled wrote:
On Wednesday 09 June 2004 19:18, Shachar Shemesh wrote:
 

Maxim Kovgan wrote:
   

man pthread_create
man pthread_mutex_init
     

Why would threads and mutexes help me at all?
   

They won't, wrong direction. Instead, why blocking SIGCHLD
and getting to possible critical races?
RTFM pselect, and more specifically, select_tut, for an answer to that 
one. The short answer is "because it's the cleanest solution, assuming 
you have a kernel implemented pselect". Unfortunetly, Linux doesn't, 
hence the trouble we're seeing.

An simpler alternative
seems to set up signal handlers for SIGALRM and SIGCHLD
(before forking), then fire alarm(2) (or ualarm(2)) and simply
sigprocmask(2) for one of them to happen?
 

I'm not sure I see how sigprocmask enters into the picture. Are you 
saying to cancel one as soon as the other happens? That sounds pretty 
certain to fail.

Still, I think this is a jackpot. Suggested pseudo code:
sigprocmask disabling SIGCHILD and SIGALRM
ask for ualarm for wakeup
fork (child starts ping, with or without cancelling ualarm first)
sigsupend on the above two signals
check which one, send INT to ping if alrm
waitpid
return mask to usual, cancel alarm, etc.
Does anyone see a fault in this setup, aside from the possibility that 
whoever called me may have had other plans for ualarm?

         Shachar
--
Shachar Shemesh
Lingnu Open Source Consulting ltd.
http://www.lingnu.com/
=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]