Updated Cray patch against openssh SNAP-20010725

[William L. Jones]

I made the change based on the following logic:

  1) If you have HAVE_SIGACTION then the signal is handler is not reset and
     thier is no need to reset it in the signal handler.
  2) If you don't have HAVE_SIGACTION then the signal handler may or may not
     be reset.  So I do the save thing a reset it.

Since SIGACTION is a POSIX standered I can depened on 1.

Case 2 insure that this code will work the same as it always does if
you don't have SIGACTION.

I really want to keep as many special out of the openbsd code as I can.

Bill Jones


> -----Original Message-----
> From: mouring at etoh.eviladmin.org [mailto:mouring at etoh.eviladmin.org]
> Sent: Thursday, July 26, 2001 10:16 AM
> To: William L. Jones
> Cc: openssh-unix-dev at mindrot.org; jones at tacc.utexas.edu
> Subject: Re: Updated Cray patch against openssh SNAP-20010725
>
>
>
> > I decided to go into a little more detail about the patches to
> > each module:
> >
> [..]
> >  serverloop.c	- Modified sigchld_handler to only reset
> > 		       SIGCLD if HAVE_SIGACTION is not set.  This is the
> > 		       only case its needs to be reset, the side effect
> >    	            is it fixes the problem with a infinite signal loop
> >  			  under unicos and unicosmk.
> Hmm.. I suggest if you want to keep doing this to do
>
> #if !defined(HAVE_SIGACTION) && defined(_CRAY)  because this could be
> valid on other platforms.  mysignal drops back to signal() if sigaction()
> can not be found (most notiable NeXTStep).
>