Bug#235241: bug resolved
On Monday 15 March 2004 03:03 am, Christian Guggenberger wrote:
> On 11.03.2004 21:41 Oliver Johns wrote:
> > On Thursday 11 March 2004 05:00 am, you wrote:
> > with 4.3.0-2.
> >
> > > So, just curious and for the record, would you please try if
> > > things also do work, when you add
> > >
> > > Option "VBERestore" "1"
> > >
> > > to your device section?
> > >
> > > thanks.
> > > - Christian
> >
> > Dear Christian,
> >
> > Thanks for the suggestion.
> >
> > If I put
> > Option "VBERestore" "1"
> > in the Device section of XF86Config-4, then I find that the
> > keypress C-A-F1 will still switch to vt1. HOWEVER, when I use
> > A-F7 to switch back, the screen goes blank.
> >
> > If I put
> > Option "VBERestore" "0"
> > then everything works totally correctly. (I assume that "0" is
> > the default.)
> >
> > I have an 845G chipset. I think that VBERestore option was put
> > in to help people with 830G chipsets. According to
> > http://www.xfree86.org/~dawes/845driver.html
> > this option was to allow 830G users to optionally back out a fix
> > that had to be put in to get correct behaviour for my chipset,
> > the 845G.
> >
> > After all of this, I think my problem was not the drivers at all.
> > It turns out that 4.3.0 changed the rules for Xmodmap. The trap
> > keys like C-A-F1 that had been implicit, were put explicitly into
> > the internal xmodmap. If you do "xmodmap -pke" as root in a
> > terminal, you'll see them in there.
> >
> > I was using an old modified Xmodmap to remap my keyboard, a
> > Happy-Hacker's model, to make the diamond keys do a
> > Delete-forward. But my Xmodmap didn't have the new entries for
> > the trap keys! So I had to make a new Xmodmap starting from a
> > current output of "xmodmap -pke". That solved the problem.
> >
> > So I don't know which Debian update "solved" my problem, since my
> > problem was of my own making and not something about the drivers.
> > Sorry. Hope the above data point helps.
>
> Thanks for the reply, Oliver!
> I was just curious, if the "VBERestore" option really was the fix
> for your VT switching problems.
> As you said, it was not, as your problems had been caused by your
> custom Xmodmap...
>
> Anyway, now we do a least know, that the VBERestore Option (set to
> false) really prevents X server crashes, while switching VT's.
> Branden you might want do undo "SVN commit: r1147 - trunk/debian"(
> the bug closer), as the change to the i830 part of the i810 driver
> has nothing to do with this Bug (IMHO).
>
> - Christian
Dear Christian,
I think that the VBERestore option was put in by Dawes because the
earlier fix for 845G chipsets like mine was making problems for older
830 chipsets. So only 830 chipset users really need to worry about
that option. It's default is "false", which is just fine for my
chipset and probably all ones later than 830.
Just for the record, in case someone out there has similar problems,
here is what happened in my case:
(1) The initial update 4.2.1-->4.3.0 caused my C-A-F1 combo to stop
working.
(2) After some futzing around, I filed a bug against xserver-xfree86.
(3) Sometime later, an incremental update of 4.3.0 was made. But, on
the same day, the update of kde-3.1 to kde-3.2 was made. The result
was that my C-A-F1 started working again.
(4) I was delighted, and filed a bug-fixed comment to my earlier bug.
(5) Then I notice that my remapping of the keyboard to make the
diamond keys act as forward-delete was now not working.
(6) I found that the kde-3.2 no longer had lines
in /etc/kde3/kdm/Xsetup to call xmodmap and remap the keyboard. So,
I added those lines back.
(7) Now I found that my diamond keys were working, but that C-A-F1 was
broken again.
(8) Finally the light dawned. I backed out the lines to call
xmodmap. Then I did xmodmap -pke > xmodmap-new. Then I made the
modifications about the diamond keys in the file xmodmap-new
and moved it to /etc/X11/Xmodmap. Then I put back the lines in
/etc/kde3/kdm/Xsetup to call xmodmap and use /etc/X11/Xmodmap as the
new xmodmap.
(9) Now everything works OK.
It is entirely possible that the INITIAL update to 4.3.0 already
solved whatever problems exist with the 845G chipset. After all, I
was able to use chvt 1 to change to vt1 and A-F7 to change back.
When I put Option "VBERestore" "true" in XF86Config-4 I do find that
switching back from vt1 to vt7 is broken. So disabling Dawes's fix
does indeed break my 845G system, so the fix is doing something for
me. I just don't know which Debain update was the first to have
that fix in it. Probably the first one, but I can't be sure.
Thanks for your patience. And I still thank you guys for the frequent
updates.
--OJ
Reply to: