Bug 371756

If just this machine (my test machine) is affected, feel free to close this with WORKSFORME. But please test with other hardware if it works at all.

We might have some bad setting in the xorg.conf we use during installation.

works with my test machine

It's impossible to install Alpha2+.

[...]
Linuxrc crashed. :-((
Press ENTER to continue.

Please reopen, if the issue still occurs with Alpha3.

Still the same with Alpha3 (as of 2008-03-18 15:50 MET).

There is no Alpha3 released yet. So how did you manage to install it?

Whatever that thing is that announces itself as "Alpha 3" via SLP.

Ok. I'll try a SLP installation then.

How can I reproduce this? I tried with german keyboard setting and opened an xterm with Ctrl-Alt-Shift-x, in which autorepeat works fine during installation. You can also check the autorepeat setting with the 'xset q' command inside a xterm.

Can you please reopen once you provide the requested feedback? Thanks.

It didn't work in any Qt input field during installation. If it works in general, that's good enough for me - the problem might as well be the cheapish keyboard of my test PC.

Where can I find a Qt input field during installation?

The user name in "create user", for example.

Indeed I can reproduce this issue. Luckily it's before the installation begins.
According to 'xset q' autorepeat is enabled. After running 

  xset r rate 660 25

which are exactly the same settings as before autorepeat works again!?! No, this doesn't make any sense. I can't explain the behaviour. Therefore if you think this is an important issue, I suggest to apply this workaround to YaST2.

See also bug #433338

*** Bug 438749 has been marked as a duplicate of this bug. ***

*** Bug 441093 has been marked as a duplicate of this bug. ***

Too many duplicates.

(In reply to comment #13 from Stefan Dirsch)
> I can't explain the behaviour. Therefore if you
> think this is an important issue, I suggest to apply this workaround to YaST2.
> 

Which workaround?

The xset command in exactly this comment #13.

Reassigning to Stano, since I can't apply the workaround.

Lukas, I'm not sure if this for you or Jiri.

cd source/installation
grep -r "xset r on" * | grep -v "\.svn"
src/include/misc.ycp:           WFM::Execute (.local.bash, "xset r on");
startup/YaST2.call:                     xset r on

There are actually two places in installation that call `xset r on`, does it mean that calling that command is not enough? And how comes that if I open an xterm and write that command (during installation), autorepeat suddenly starts working?

I have no idea, but maybe it's enough to run it each time you changed the keyboard layout. Just a wild guess.

In that case it actually belongs to keyboard setup...

Stefan, this bug is marked for 11.0, did you reproduce it for 11.1?

And when Lukas mentions xset command, it is a different one from your proposal in comment 13. So if YaST during start (in YaST2.call) would call "xset r rate 660 25", it may work regardless the change of the layout... Did anybody test it?

Apparently the bug has been reproduced on 11.1 by a lot of users. Nobody tested
if "xset r rate 660 25" or "xset r on" helps in all cases. It's just wild guessing that it will help after changing keyboard layout.

The problem is what should help: there is some current code with "xset r on", but nothing with "xset r rate 660 25". I don't know why we should have more such commands at more places, so I am asking.

If I create a patch for you, that calls "xset r rate 660 25" after setxkbmap call, are you able to test it that it helps?

All the reporters, i.e. the initial reporter and all the others which bugreport
for 11.1 has been closed as duplicate of this one should be able to test if it works, yes.

Well, I was asking if you can test it, not the reporters.

yast2-country-2.17.24 contains "xset r rate 660 25" after each setxkbmap call during installation.

As this can't be said to be fixed, moving to Stefan, which should find the real source of problems.

The real source of the problem is that X Input in X.Org has changed a lot since openSUSE 10.3, i.e. with xorg-server 1.4 and I doubt that this issue can be fixed with reasonable efforts.

Thus WONTFIX.

Hey, how could it be WONTFIX if there is a real bug? Can't you escalate it upstream?

The problem occurs while going from the 'Welcome' YaST screen to the
'Installation Mode' screen when the user hasn't pressed any key on the
keyboard, yet. Pressing any key (even if it doesn't trigger any action) will 
result in autorepeat on.

Apparently YaST loads the keyboard layout somewhere during the transition   
between the two screens (probably while doing system probing).

The reason for this is the changed default handling of the core keyboard:   
the core  keyboard is identified and the keyboard state is loaded on the    
first key press by the user. If the keymap gets modified before this is done 
the autorepeat state may not be the one of the core keyboard.
Thus this is not a bug but a change in the semantics which was introduced   
with dynamic input devices and the virtual keyboard.

I remember this being discussed on the last XDS in Edinburgh (maybe Luc also 
remembers).

The situation can be remedied by just calling 'xon r on' or just by calling 
XChangeKeyboardControl(3) with the appropriate arguments after the keyboard 
layout has been loaded to turn on autorepeat.  XAutoRepeatOn(3) may also    
already be sufficient. So please fix YaST.

'xon r on' is called by YaST, see some previous comment by Lukas...

Then it's most likely done at the wrong place.

Here is what I've tested:
- Started installation from CD (Beta4)
- After YaST has started:
  - switch to vc2
  - edited /etc/ssh/sshd_config to allow password logins
  - set a root password
  - did: ifconfig eth0 up
  - did: dhcpcd eth0
  - ran: rcsshd start
  - logged in from remote
(1) - on vc2 ran /bin/sh /usr/lib/YaST2/startup/YaST.call installation inital
(2) - navigated thru welcome screen purely by mouse
    - when 'mode' screen showed did: 'xset r on -display :0; xterm -display :0;
    input in xterm had autorepeat.
  - redid step (1) and (2)
  - when 'mode' screen showed did: xterm -display :0;
    input in xterm had *NO* autorepeat.

So it is possible to turn it on doing 'xset r on'.
Since I don't know what actions you are performing in YaST I cannot tell you where to turn it on. Please check your YaST command sequences and find out where you have to enable autorepeat.

I think Egbert means 'xset r on'. Still interesting news. Never new about this change. Maybe we can also get rid of this stupid workaround to reload the keyboard layout in /etc/X11/xinit/xinitrc.common to make keyboard layout working also with kdm/gdm autologin.

OK, I'll try to play around this again, but after SLE11 is released. Now when we have working solution it's better to stay with it.

I'm not sure why the current placing of 'xset r on' is wrong (now it is at the very beginning), but according to the previous comments, it should be correct to call it after all the keyboard settings are done, i.e. just before the client with welcome screen exits. And remove those "xset r rate 660 25" calls after each setxkbmap.

*** Bug 484211 has been marked as a duplicate of this bug. ***

OK, now (from yast2-country-2.18.4), country calls just 'xset r on' after the keyboard is set (which is at the end of initial screen).