Fix for call-ar-st.exp failure on x86

[Jim Blandy]

> > If you see anything dumb, let me know.
> 
> Well, since you mention it :-).  Here is a fix to an uninitialized
> variable:
> 
> 1999-10-12  Jim Kingdon  < http://developer.redhat.com >
> 
> 	* valops.c (value_push): Initialize container_len even if
> 	PARM_BOUNDARY is zero.

Wonderful!  Programmers are supposed to write a bug every four or five
lines of code on average, so we're almost up to quota by now --- soon
this code will be bug-free!

*bleah*