Re: Native build...

[Johnny Billquist <bqt%softjar.se@localhost>]

Greg Oster wrote:
> hbent%cs.oberlin.edu@localhost writes:
> > On Wed, November 26, 2008 3:26 pm, Johnny Billquist wrote:
> > > hbent%cs.oberlin.edu@localhost wrote:
> > > > On Wed, November 26, 2008 11:39 am, Johnny Billquist wrote:
> > > >
> > > > > pic: Unsupported relocation type 168 in non-PLT relocations
> > > > >
> > > > > grops: Unsupported relocation type 136 in non-PLT relocations
> > > > I've seen these "unsupported relocation type" errors when trying to run
> > > > large cross-compiled programs.  I figured there was a problem with my
> > > > cross-compiling setup, but perhaps that's not the case.
> > > Nope. That's a problem that have been around for a long time (atleast
> > > four years, and probably longer. I can't even remember anymore).
> > > The only places I've seen it is with groff.
> > > I've suspected something in c++ for a long time, but I plainly can't get
> > > my head around gcc (a combination of too little time also contributes...).
> > I'm seeing it in glib 2.18.2 which doesn't have any c++ code.  Running
> > gtester gives "gtester: Unsupported relocation type 232 in non-PLT
> > relocations" and gobject-query gives "gobject-query: Unsupported
> > relocation type 72 in non-PLT relocations".
> >
> > I am using netbsd's patched gcc 4.1.3 with stock binutils 2.19.  Perhaps I
> > will switch to netbsd's binutils to see if that makes a difference.
>
> Could folks please add these sorts of additional details to this PR:
>
>  http://gnats.netbsd.org/39182
>
> Hopefully that'll help whoever ends up looking at the issue... 
>
> Thanks...
>
> (And yes, I'd like to see vax self-hosting for 5.0... right now a 
> cross-built netbsd-5 for vax can't build itself natively...)
>
> Later...

It's even worse.
I tried booting 5.0_BETA on an 8650 today, and it don't work at all.
The code that detects the hardware have broken. Looking at the 
repository, I think that Matt's "rototoll of the vax" could be the cause.
If nothing else, his changes brought NetBSD back into a state of 
mismatch with reality with regards to the 86x0 machines, since he 
removed the Abus stuff that I clubbed together many many moons ago.
(NetBSD now once more thinks that the SBI is the main bus of an 86x0 
machine, which it isn't.)
Sigh!

Anyway. Booting 5.0 on the 8650 fails to find the UDA50 on which the 
system disk sits, which is on the first Unibus of the machine (yes, it 
has more than one Unibus).
So, obviously, I can't get any further.

        Johnny