syscall error, was Re: [ntp:hackers] Solaris still a problem

gnu not unix gnu at wraith.sf.ca.us
Fri Jan 9 18:58:25 PST 2004


In message <4.3.1.2.20040109210844.00e18360 at pop.gis.net> you write:

>At 06:06 PM 1/9/04, gnu not unix wrote:

>>I can trigger the "interrupted system call" message at any time, simply
  (...) 
>>addto_syslog: select(): nfound=-1, error: Interrupted system call

>I've seen that regularly. It's in ntpd.c and happens if debug > 2. I don't think
>we should be logging EINTR errors at all, since that's not unexpected in
>the select (The error handling code above the logging checks for that errno).
>I'd like to turn it off unless someone thinks it's really, really important.

Hi Danny--

I agree, this EINTR should be hushed up. The tone of the message sounds
ominous but in this case looks to be spurious in the actual code, as
you see. 

Harlan, are you in the "want to see every possible error return"
defensive programming mode? In my somewhat casual socket programming,
this "error" has only been something that I've ignored in my code.

I suppose setting the debug to "highest" should be ok. By the way, 
where is that list of what the suggested debug levels are, and what
value "highest" is? I sort of recall seeing this somewhere. Thanks.

../Steven




More information about the hackers mailing list