Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!tut.cis.ohio-state.edu!ucsd!hub!tweedledee!steve
From: steve@tweedledee.uucp (Steve Trainoff)
Newsgroups: comp.sys.next
Subject: Re: rrn on NeXT
Summary: Problem with gcc Optimiser
Keywords: gcc Optimiser bug?
Message-ID: <4133@hub.UUCP>
Date: 2 Mar 90 02:23:21 GMT
References: <4568@accuvax.nwu.edu>
Sender: news@hub.UUCP
Reply-To: steve@tweedledee.ucsb.edu (Steve Trainoff)
Organization: University of California at Santa Barbara, Physics program
Lines: 44

In article <4568@accuvax.nwu.edu> cliff@delta.eecs.nwu.edu (Cliff Chaput) writes:
>Has anybody gotten rrn (remote rn that accesses the NNTP server on another
>machine, not real rn) to run on their next?  I have a hacked version, and I
>can read news just fine, but when I attempt to post I occasionally get an
>error which forces rrn to exit: %interp buffer overflow.  If you've gotten
>rrn to work on your NeXT, please let me know.
>
>Thanks a lot,
>
>Cliff Chaput
>Mneme Project - NU Psychology Dept.
>cliff@mneme.psych.nwu.edu, cliff@eecs.nwu.edu

I too have had this problem, and the resolution was rather interesting.
I recomplied the code with the debugger and ran the program under
gdb.  I found the offending code.  It was in intrp.c:

	 register  char *dest;
	...
	switch (*++pattern) {
	...
	default:
	    if (--destsize <= 0)
		abort_interp();                   <- Generates the error msg
		*dest++ = *pattern | metabit;     <-- The problem!
		s = nullstr;
		break;


I found that despite the code written, dest was NOT being incremented.
When I turned off the -O switch on the compilation, the problem went away.
I wasn't able to reproduce the problem in a simple test case.  
The temporary soln is to compile this code separately by putting the
following lines into the makefile.

# There seems to be a problem with the gcc optimizer that manifests itself in intrp.c
intrp.o: intrp.c
	$(CC) -c -g intrp.c 
	



..STeve
steve@tweedledee.ucsb.edu (NeXT mail)