Subject: Can't build GCC-2.7.2 under 2.4.5!!!
To: None <amiga@NetBSD.ORG, current-users@NetBSD.ORG>
From: Jake Hamby <>
List: amiga
Date: 12/03/1995 12:00:41
Well, my patience in waiting for NetBSD-1.1 paid off, and I now have a 
working NetBSD-Amiga box to complement my FreeBSD-i386 box.  I had a bit 
of trouble installing it to a Zip disk (which wasn't really big enough 
anyway!) but after buying a new Quantum 730MB SCSI drive, I installed 
it okay, and now it works perfectly, except for one problem...

I'm a little concerned (and have been) about the OLD version of GCC 
(2.4.5) being used for NetBSD.  I understand that there are some problems 
with versions such as 2.6.3 on various architectures, but hopefully, now 
that GCC 2.7.2 is released, it has fixed up any remaining concerns (I 
know there were a bunch of m68k patches between 2.7.0 and 2.7.1).  At any 
rate, having built GCC a number of times in the past, I assumed it would 
be pretty easy to build GCC 2.7.2 on top of the GCC 2.4.5 on my 
NetBSD-Amiga box.  Not so!  I've had at least three problems, so far (and 
still not finished), two of which I'm going to have to use my FreeBSD box 
to bail me out with:

1) Bison didn't work properly when compiled with -O optimization, it 
produced spurious "0377" characters (that's 255 in decimal, of course).  
It looks like a y with two dots over it, BTW...  When compiled with -O2 
or no optimization, it worked okay.  Normally you don't need bison to 
build GCC, except that I had started with 2.7.0 and applied patches to 
bring it to 2.7.2 so some of the source ".y" files got updated 
necessitating a rebuild of the '.c' files.

2) In the process of building GCC, it builds a number of temporary 
programs (genoutput, genattr, gencodes, etc) to parse a .md file (in this 
case and generate various .c files that are then compiled.  When 
it compiled, genoutput, the file generated (insn-output.c) again had 
spurious "0377" characters, no matter what level of optimization I 
compiled genoutput with, it had the same problem.

3) Now, another temporary program "bi-arity" is simply core dumping 
instead of generating its output file.  In both this case and the last, 
I'm going to have to use FreeBSD to bail it out, and generate the correct 
file, but I'm afraid that if there are this many bugs in programs 
generated by this version of GCC, will the stage1 compiler be bug-free 
enough to generate the final version of GCC???

Has anyone successfully built GCC-2.7.x on the m68k platform yet?

     Jake Hamby                         |   E-Mail:
  Student, Cal Poly University, Pomona  |   System Administrator, JPL