Re: setsid

[Bruno H. <br...@cl...>]

Sam wrote:
> > Not sure whether to comment it for pipes or #ifdef the SETSID() to do  
> > nothing in MT?
> 
> I don't know anything about these system calls.

setsid:
  http://www.opengroup.org/onlinepubs/9699919799/functions/setsid.html
setpgrp:
  http://www.opengroup.org/onlinepubs/9699919799/functions/setpgrp.html
setpgid:
  http://www.opengroup.org/onlinepubs/9699919799/functions/setpgid.html

> if you know what you are doing and are confident that setsid does not serve any 
> useful purpose, you can remove the calls outright (with an extensive 
> explanation in the changelog).
> otherwise, you could disable it just in MT builds.
> if you want to be extra careful, you can write an autoconf test to detect 
> whether setsid can be used the way we do, and disable it if it cannot be.
> Bruno, can you comment on the matter?

There are situations where you need it, and there are situations where it
hurts to have SETSID() called in the freshly created subprocess. Probably
I encountered a case where it was needed, years ago, and put in the call,
so that things behaved "correctly" w.r.t. Ctrl-C and logout. Maybe this case
was that when the user presses Ctrl-C, the user expects to stop clisp's
execution, but without the SETSID() the system sends a SIGINT signal to the
subprocess, thus killing it. (Because clisp's signal handler for Ctrl-C has
no effect on the subprocess.)

Bruno