On Sun, 15 Apr 2007, Johnny Billquist wrote:

> Just tried your suggestion of building gcc with -O1, and no change for me.
> Tried to compile a small program of my own with -O1 and -O2 and both crash 
> with internal compiler failure.
>
> Are you actually having success with these patches?

   I had thought so, but I was also trying several other things, and I may 
have been using a cc1 and cc1plus generated differently.

   That said, I started a build yesterday with Matt's change to default to 
-O on the vax, and got to x11/bin/lbxproxy before it quit with a segment 
violation compiling lbproxy/os/connection.c.  I haven't started looking at 
that problem yet, but I did copy the usr/libexec/cc1{,plus} that build 
created to my vax.  I was able to get a large part of a tools build run 
until it quit while running makeinfo during the gcc4 build.  I expect the 
update rerun of the tools build to fail twice more while trying to 
generate a couple of .ps files, and then I will run a kernel build using 
those tools.

   I suspect that if someone digs deep enough into these failures and can 
locate the (presumably) bad code generated, Matt may be able to fix it. 
I've been concentrating trying to get to the point where I can do a tools 
and kernel build on the vax, and then to try getting the appropiate 
changes pulled up to the netbsd-4 branch so that the vax will be usable 
with a 4.0 release.  Then I will probably try to dig deeper into the 
problem of building gcc with -O2.


--
Michael L. Hitch			mhitch@montana.edu
Computer Consultant
Information Technology Center
Montana State University	Bozeman, MT	USA