Subject: Re: m68k optimizer bug in egcs
To: Chas Williams <chas@cmf.nrl.navy.mil>
From: Erik Bertelsen <erik@mediator.uni-c.dk>
List: tech-toolchain
Date: 11/17/1998 20:15:10
On Mon, Nov 16, 1998 at 10:42:29PM -0500, Chas Williams wrote:
> i rebuilt egcs after the latest round of changes, but the optimizer bug
> seen on the m68k (especially when building libgnumalloc) is still there.
> 
...
> 
> however, i had a bit of time to play around with some of the development
> releases of egcs, and saw the same optimizer bug in the 19980906 release
> but by the 19980914 release, it seemed to be fixed.  the following diff
> to global.c seems to fix the problem (i.e. malloc() again only had a single
> psuedo register again).  i have applied it to the egcs in the dist/gcc tree,
> and now i can building a working libgnumalloc on m68k-netbsd w/ -O.  the
> ChangeLog entry that seems to go with this says:
> 
> Sat Sep  5 16:34:34 EDT 1998  John Wehle  (john@feith.com)
> 

I've applied the patch from the cited message, built and installed
egcs, then rebuilt (with clean) and installed libgnummalloc.

With this setup, I've done a complete kernel compilation and rebooted
this kernel.

It is now running with this kernel and happily recompiling libc af-
ter today's sup.

Unless something else pops up, it seems that the last egcs 1.1.1 related
problem that I've met on the mac68k can be solved.

regards
Erik Bertelsen