httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Bruce Korb <>
Subject Re: Adverse behavior from SIG_IGN for SIGHUP and SIGCLD???
Date Thu, 23 Jan 2003 20:42:12 GMT
T Ford wrote:
> What adverse behavior if any will I see on linux if I have a module that forks, and immediately
calls signal(SIGHUP, SIG_IGN) and signal(SIGCLD, SIG_IGN) when using the worker MPM?

It will be propagated to everything you spawn off.
Most programs presume that they can catch children
with wait() calls without having to do anything special.
With ``signal(SIGCLD, SIG_IGN)'' they are in for a
surprise.  POSIX has changed the spec for exec(2) to
allow applications (operating systems) to reset
the signal SIGCHLD (SIGCLD) specifically to SIG_DFLT,
but you cannot rely on it.  Instead, do this for every
program that needs to ignore SIGCHLD:

  void ignore_sig() {}

   signal(SIGCHLD, ignore_sig);

For the history, go back 20 years to the BSD vs. SysV wars.

View raw message