I was able to do a bit more with it now.

The Matrox still won't fly.  But the S3 ViRGE looks great, and amiwm is a sponge-solid
window manager, much better than twm.  One nice test was lavaps, an app that sends the
NetBSD-amiga X server into la-la land.  The amiwm/xfree4-alpha combo fetched and
compiled tcl, tk, and lavaps, and ran lavaps...  whoops, looks like a pkg bug:  it's
doing the same as the nb-amiga one did.  Anyway, real work can be done, if you're
careful.

I can invoke PLT symbol errors by trying to drag the screen (amiwm...  cool...)
multiple times, but any other window op seems (relatively) stable.  As long as you
have a backdoor open to the machine, you can restart xfree any number of times; but
you'll never get your console back on the local box.

Just a sample of the PLT's:

main program: Undefined PLT symbol "__divlu" (reloc type = 26, symnum = 231)
main program: Undefined PLT symbol "select" (reloc type = 26, symnum = 114)
main program: Undefined PLT symbol "__remlu" (reloc type = 26, symnum = 163)

Here are a couple of pointers to the PLT problem (do a search on mail-index, and read
the book.  It's a bugbear).  I'm new to the whole toolchain thing, but it has to do
with something lurking in the 64-bit ELF format.  First, there's this post from Steven
J. Dovich and a confirmation on its resolution from H. Feyrer (as it applied to gimp):

http://mail-index.netbsd.org/tech-pkg/2000/03/02/0002.html

I haven't had time yet to compile xfree 4 with the new flags; if you can go faster
than my Alphastation, yer welcome to try.

Second, there's this post from Paul Mather.  Dunno if it works or not; I could have
sworn I had these compats defined.

http://mail-index.netbsd.org/port-alpha/2000/10/18/0004.html

I'll be churning out some new code in the next day or so, and test away.  There's
gotta be a magic link flag out there somewhere.

--David

Andrew Hobgood wrote:

> > errors in xc (lots of the standard XFree86 warnings, but no errors).  But I
> > don't even get as far as you do, on the Matrox; the screenmode change never
> > occurs, and my console is garbaged in a most colorful way.  External net
> > services still work, though.
>
> mmm.  garbaged console.  pretty, isn't it? =)
>
> > I also have an S3 ViRGE VX (old Diamond card), and X starts up just fine, with
> > the same codebase.  But I can only do a certain number of blits-- window moves
> > or resizes-- before it locks up.  No zap keys, but only the X process is locked;
> > external services work fine.
>
> Have you tried starting X from another tty (ie, serial console or via an
> ssh session) so you can catch errors echoed to the tty after X has
> started?  I'd bet that you've got the same PLT errors that I've been
> getting (and the other Andrew, too)... I've found that ion (search on
> freshmeat.net) works great for text windows, and can remain running for as
> long as a couple of minutes before seizing, as long as I don't attempt to
> run anything other than rxvt.
>
> > Thanks all, esp. Simon; this is a mighty effort.
>
> *definitely* agreed.  Simon, the work you've done is impressive -- the
> fact that we've got some showstoppers shouldn't be allowed to overshadow
> the incredible fact that some of us can actually get the display
> *working*, which is an incredible step forward.
>
> FWIW, I've moved up to -current on my kernel (yesterday's CVS update), and
> still no luck.
>
> millennium hand and shrimp,
>
> /Andrew