[Nagiosplug-devel] check_dns: DNS WARNING - nslookup returned error status

Andreas Ericsson ae at op5.se
Fri Sep 2 06:46:05 CEST 2005


sean finney wrote:
> (sending a second copy to the list, the first got caught with too many
>  cc's)
> 
> On Fri, Sep 02, 2005 at 02:23:16PM +0200, Peter Pramberger wrote:
> 
>>Please keep in mind that ECHILD is an error condition - there was no process
>>found to check for. status' value is not defined after the call to waitpid, so
>>the result of WIFEXITED is unpredictable.
> 
> 
> yes, this is why simply changing to the switch statement proposed by dag
> would not work.  the question is, how do we get at the status in this
> case?  
> 
> this is from r'ing tfm for sigaction(2):
> 
>        POSIX.1-1990 disallowed setting the  action  for  SIGCHLD  to SIG_IGN.
>        POSIX.1-2001  allows  this possibility, so that ignoring SIGCHLD can be
>        used to prevent the creation of zombies (see  wait(2)).  Nevertheless,
>        the historical BSD and System V behaviours for ignoring SIGCHLD differ,
>        so that the only completely portable method of ensuring that terminated
>        children  do not become zombies is to catch the SIGCHLD signal and per-
>        form a wait(2) or similar.
> 
> so what, redhat is now defaulting to having that set?  i think i see a
> way around this, but i'm not sure:
> 
> - set a signal handler for SIGCHLD
> - when SIGCHLD is thrown, we do exactly the opposite from the above, and
>   *not* call any wait functions.
> 

I wrote a test-program for this. On FC3 with linux-2.6.12-1372 (weird 
patch-number, I know), when SIGCHLD is explicitly set to SIG_IGN, the 
children exit immediately and nothing good ever comes of it. When no 
sighandlers are installed, children hang around for the waitpid(2) call.

Setting SIGCHLD to SIG_IGN is the default, according to signal(7)

> i think that should leave the child as a zombie, for our later call to
> waitpid.
> 

In theory, it should work. In reality, it doesn't (FC3). I'm not sure 
why yet.

If anyone's interested in the test-program (some non-working code is 
commented out), email me off-list and I'll send it to you. oss.op5.se is 
currently down, or I'd post it there (some jackass hit a cable with a 4 
tonne construction machine).

-- 
Andreas Ericsson                   andreas.ericsson at op5.se
OP5 AB                             www.op5.se
Lead Developer




More information about the Devel mailing list