Re: XFree86 correct from source?
On Mon, Aug 28, 2000 at 12:47:26PM +0100, Chris Rutter wrote:
> I recently just built `xfree86-1' (3.3.6-10), and apart from its taking
> three days and 810MB of disk space, it apparently worked. Does XFree86
> really auto-build successfully from package source these days?
I found that it built fine from package source (albeit slowly)
However, I've had problems with the fbdev driver and the Acorn framebuffer
device. With text acceleration turned on (the default) the Acorn framebuffer
driver allocates the largest virtual y resolution if can to give some degree
of "hardware" scrolling. When probing modes in the modelist the xfree86
fbdev doesn't turn off kernel driver text acceleration (which Russell King
(author of the Acorn framebuffer driver) so when XFree86 tries a mode
(eg 800x600, virtual 800x600) the kernel driver gives it a massive virtual y,
which causes xfree86 to reject the mode as the parameters don't match.
fbdev does turn off text acceleration once it goes for the mode for real.
However, I found I needed to patch it to turn of text acceleration when
probing to make modes not get rejected. I'm not sure if it's the
right thing to do.
*** xc/programs/Xserver/hw/xfree68/fbdev/fbdev.c.orig Wed Jul 19 11:54:40 2000
--- xc/programs/Xserver/hw/xfree68/fbdev/fbdev.c Wed Jul 19 16:09:17 2000
*** 1424,1429 ****
--- 1424,1431 ----
var1.yoffset = checkframe_y(fbdevInfoRec.frameY0, &fb_fix, &var1);
var2 = var1;
var1.activate = FB_ACTIVATE_TEST;
+ var1.accel_flags &= ~FB_ACCELF_TEXT;
+ /* Turn of text acceleration; no use to us, causes problems on acornfb */
if (ioctl(fb_fd, FBIOPUT_VSCREENINFO, &var1))
if (var1.xres != var2.xres || var1.yres != var2.yres ||