I don't disagree with Harlan about stepping back and understanding what needs 
to happen. But there is a real issue with sntp, atleast on some systems, 
waiting forever. It happens because the setjmp is not saving the signal masks - 
at leat on linux. This is the behavior if the sntp server never responds:
  
>in read_socket
>setting up alarm for 3 secs
>Before recvfrom
>in jump_handler
>Time out occurred
>in read_socket
>setting up alarm for 3 secs
>Before recvfrom
--------------WAITs forever
   
  Linux documentaion for setjmp recommends the use of sigsetjmp, and patch I 
mailed in the previous mail addresses this. BTW, internet.c also suffer from 
same issue.
  -Anil
   
   
  
Harlan Stenn <[EMAIL PROTECTED]> wrote:
  I think you may benefit from taking a step back and understanding exactly
what you want to have happen during startup.

If all goes well, sntp will bet a response "quickly".

If not, how long do you want to wait for an answer?

What do you want to do if an answer does not arrive?

It may be that you want to avoid using sntp (or ntpdate) entirely, and start
ntpd -g as early as possible, using iburst and a usefully persistent drift
file. You might also be interested in:

https://ntp.isc.org/bugs/show_bug.cgi?id=472

Anyway, consider starting ntpd "early" and then continue with your startup
and then check to see how well things are going using, for example, the
ntp-wait script to make sure things are OK before "going live".

H

_______________________________________________
questions mailing list
[email protected]
https://lists.ntp.isc.org/mailman/listinfo/questions
  


                
---------------------------------
 Yahoo! Personals
 Skip the bars and set-ups and start using Yahoo! Personals for free
_______________________________________________
questions mailing list
[email protected]
https://lists.ntp.isc.org/mailman/listinfo/questions

Reply via email to