Subject: Xfree86 using Simon's binaries, a report
To: None <email@example.com>
From: Andrew van der Stock <firstname.lastname@example.org>
Date: 01/21/2001 01:13:10
I compiled up a 1.5Q kernel from GENERIC with INSECURE set and just enough
devices to boot my PC164. That's fine, it reboots fine and takes 2 MB less
disk space than GENERIC. Funnily enough, I have to link in apecsa and pcmcia
to link the kernel, which means that there's something not paying attention
to the config files in these ares.
I downloaded the binary snapshot and untarred it over the top of what I had
before; Simon's snapshot completely replaced my last build, which is not
surprising as the my 4.0.2 and Simon's patches were very nearly one and the
same as my patches were based upon Simon's 4.0.1 work.
The Xserver crashes after a few seconds, usually with a PLT error on
something like __errno, memcpy, or free. I get this reliably by getting one
of the twm xterm windows that is partially obscured and resizing and forcing
it on top in a few mouse clicks. It's safe enough for a little while if you
use a window you don't have to expose. The console is frozen then in
graphical mode. It's possible to restart X from a ssh session on another
box. After a crash, none of the keys that perform mode changes (ctrl-alt-+
or -) or ctrl-alt-backspace work, which is expected when there's no Xserver
to intercept the keystrokes. However, wscons doesn't re-appropriate the
console. Haven't bt'd a core dump or a ktrace yet.
A sorta TODO list, of which I'll get around to unless someone beats me to
If wscons detects that the process that took control of the wsconsole has
died, it should reset it to a sane state.
The Xserver cannot take over the console unless you are logged on as root on
The Xserver does not use the i2c DDC info to fill in the h and v max limits;
I had to specify those even though it is able to detect my monitor's
settings on that one.
I had to put in a couple of extra lines in the keyboard section to make it
use wskbd and /dev/wskbd0 or otherwise just typing a couple of characters
would cause the Xserver to quit.
If I boot with the KVM switch pointing away from the alpha, the mouse does
not get detected and X fails to start. The Xserver should ask wsmouse to
re-detect the mouse if it's not detected.
Building KDE 2.1 beta 1
- Building QT 2.2.3 - wants to link with /usr/lib/c++crt0.so and this is a
Sparc only package, according to pkgsrc. WTF?
basically the same as Simon's, except that I'm using a Millennium I. This is
the same driver as the G200, and as one of the authors of that driver, the
major differences between the G200 and the Mill I is initialization, where
the RAMDAC is (the 3026 is a separate chip in the Mill 1 and lower in spec),
and if there's a 4 MB barrier limitation (which is easily worked around if
you use modes that are not 4 MB in size, like the modes I use), and a couple
of very very minor acceleration differences (the G200 2D GPU is all but
identical except is has a couple fewer limitations on scr-scr copies (which
are limited to roughly 256 kb on the Mill I/II, which I don't know that we
make use of anyway).
I'm running a very plain 1.5 system, that is starting to get a few packages
like CVS and Perl and so on now as things like automake and autoconf become
needed. I have not updated userland nor bootblocks to -current, just the
Andrew van der Stock, email@example.com http://www.greebo.net
SAGE-AU President http://www.sage-au.org.au