Subject: Re: Three X server problems on NetBSD/shark 4.0
To: Ignatios Souvatzis <email@example.com>
From: Michael Lorenz <firstname.lastname@example.org>
Date: 12/28/2007 13:11:35
-----BEGIN PGP SIGNED MESSAGE-----
On Dec 28, 2007, at 08:23, Ignatios Souvatzis wrote:
> I've run the X server on isgfb with the (appended) XF86Config, and
> have some complaints:
> C1. The X cursor is somewhat inside-out - I didn't yet count pixels,
> but it looks as if the bytes of the bitmask are swapped. E.g., the
> standard X cursor doesn't look like >< , but like <>.
Gah, I thought I committed the fix for that. Gimme a day or two...
X seems to make funny assumptions about bit order in cursor images at
times, igsfb is supposed to deal with that on both shark and krups
but apparently I forgot something there :/
> Q1: (Some?) X servers allow to force a software rendered X cursor;
> can I do this with the Shark XFree86, to check whether the problem
> is in the hardware cursor code?
Option "HWcursor" "no"
in your device section.
> C2. Much worse: sometimes key release events get lost. It's mostly
> harmless when text keys are affected, because pressing it again
> will stop the auto-repeat, but I've had the caps lock or even some
> sort of control or meta key event (couldn't find out) block in a
> way that prevented me from using the keyboard at all in a controlled
That's weird, I don't remember that kind of problem. Did anything in
ARM interrupt handling change?
> C3. I can't switch back to wscons at all, and when trying to force
> 16bpp mode, I've had the keyboard still drive the text console (and
> scroll up the X part of the framebuffer).
Switching in and out of X never worked right, shouldn't be enabled at
all. And 16bit mode won't work either unless you have patches rotting
in my source tree for a while now.
Besides that - weird things may happen if X doesn't start on virtual
> The build machine was NetBSD/amd64-4.0_RC5, in case this matters, for
> the kernel as well as the userland.
Shouldn't matter, I built and used shark kernels and userlands on
FreeBSD/i386, macppc and sparc64 - no problems at all.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)
-----END PGP SIGNATURE-----