Bug#301259: 2.6 i386 installs broken with French keyboard (and possibly others)...
severity 301259 important
(maybe even serious if we consider that breaking 2.6 installs is RC)
I'm afraid I have to confirm this bug report. Of course, this is too
late now. I *should* have tried to reproduce this bug with a 2.6
install while I only tried with a 2.4 install which works well.
In 2.6 installs with fr-latin9 keymap (ie the "French" choice in
kbd-chooser), the installed keymap fails to load at boot time in 2nd
stage (and on later reboots).
This may be fixed after the install by running "install-keymap
fr-latin9" or by running "dpkg-reconfigure console-data".
2.4 installs are not affected by this bug.
See the IRC discussions about this:
14:12 < bubulle> dammit, a RC3 netinst CD fails to load the French keymap....
14:13 < streuner> how do i get a shell with d-i?
14:13 < streuner> at boot-time?
14:13 < fjp> bubulle: Red screen error?
14:13 < bubulle> fjp: no, in 2nd stage
14:13 < vorlon> streuner: ALT-F2?
14:14 < bubulle> fjp: I saw some kind of error about "install-keymap" during the boot.
14:14 < streuner> vorlon: thanks
14:14 < bubulle> I have completed the install and I'll reboot just to see what the error may be
14:14 < vorlon> streuner: and ALT-F1 to get back
14:15 < fjp> bubulle: Enable bootlogd before you reboot
14:15 < bubulle> fjp: I will....at next boot..:-)
14:15 < bubulle> looks like an error in the keymap
14:16 < bubulle> this is a 2.6 install
14:16 < bubulle> null symbol found (6 times)
14:17 < bubulle> unknwon keysymp 'kappa'
14:17 < bubulle> /stc/console/boottime.kmap.gz: 992: syntax error
14:17 < bubulle> syntax error in map file
14:17 < bubulle> key bindings not changed
14:18 < fjp> Weird that you have not seen this before...
14:19 < bubulle> fjp: well, I give netinst CD's less testing now as I can't boot them in vmware...:-(
14:20 < fjp> but how can it be different from netboot in this respect?
14:20 < fjp> bubulle: Use qemu ;-)
14:20 < bubulle> fjp: I should probably...just need the time for making a correct setup
14:20 < Kamion> does anyone know a good reason for partman-auto's recipes to bother setting the root partition to be bootable?
14:21 < Kamion> it breaks Windows when you're doing guided partitioning on just the free space, rather than on a whole drive, you see ...
14:22 < fjp> Kamion: I doubt it is needed if you partition the whole drive and install bootloader in mbr.
14:23 < eugen> bubulle: it is probably the same error as for trq|trf|hebrew|lt|lv-latin4|ua in kbd-chooser. See it's prebaseconfig
14:23 < vorlon> right; if you don't install to the mbr for some reason, you either have to install to a bootable partition, or have to have some other bootloader that chainloads you
14:25 < eugen> bubulle: what keymap? fr?
14:25 < bubulle> eugen: right "install-keymap fr-latin9" fixes the problem
14:27 < bubulle> this does not happen with businesscard CD's.
14:28 < Kamion> vorlon: it seems to me that lilo-installer ought to handle that in the case when you're installing to a bootable partition
14:28 < bubulle> well, IIRC, at least. During the test phase, I DID manual fr installs with businesscard CDs
14:28 < fjp> bubulle: Which images are you using: if you are installing everything from testing it should not make any difference
14:28 < bubulle> fjp: yep, I know...this is what puzzles me..:-)
14:29 < Kamion> - Some BIOSes don't properly recognize disks with an unusual partition
14:29 < Kamion> table (e.g. without any partition marked active) and refuse to boot
14:29 < Kamion> from them.
14:29 < Kamion> hmm, that too
14:30 < vorlon> ah, indeed
14:30 * bubulle burns a RC3 businesscard and ries immediately
14:32 * eugen hates azerty keymaps :)
14:33 < eugen> it hard to enter passwords with them :)
14:33 < bubulle> eugen: that's on purpose..:-)
14:33 < bubulle> eugen: ask vorlon. He *loves* azerty keyboards
14:34 * bubulle tries running qemu...
14:34 < Kamion> actually lilo-installer already offers to make a partition active if you're installing to one
14:34 < vorlon> bubulle: no, I love dvorqk keyboards ;)
14:35 < Kamion> so the only thing remaining, I think, is to arrange for grub-installer and lilo-installer to make the boot partition active if there are *no* active partitions (since apparently that state can break some BIOSes)
14:36 < bubulle> fjp: do I have to create the file I'll use as hard disk with qemu?
14:37 < eugen> bubulle: make it with dd
14:40 < bubulle> wow, qemu runs damn fast
14:41 < bubulle> hmm, not that fast after all..:-)
14:41 < bubulle> warning: could not open /dev/net/tun: no virtual network emulation
14:41 < bubulle> I still have to figure out how to have networking work
14:42 < fjp> bubulle: Let me send you the file I use for networking
14:44 < eugen> bubulle: I reproduced the problem on the running system (sid), but I have "unknown keysym 'dagger'"
14:45 < fjp> eugen,bubulle: maybe it is some other package that provides the keysyms that has broken
14:45 < bubulle> OK....problem reproduced also with businesscard CD...:-(
14:45 < bubulle> I even get a kernel panic....
14:47 < eugen> fjp: no, that the way how kbd-chooser installs keymaps is broken
14:48 < bubulle> eugen: so, RC bug against kbd-chooser?
14:48 < bubulle> This is #301259, indeed
14:50 < bubulle> the bug reporter didn't answer my request about the installation type (2.4 or 2.6)
14:51 * eugen wonders how many keymaps do not work in second stage?
14:53 * bubulle runs a 2.4 install on the same machine
14:54 < fjp> eugen: So how is it wrong?
14:54 < bubulle> fjp: oh, thanks BTW for the qemu setup file.
14:54 < bubulle> I think I'll soon drop vmware..:-)
14:54 < fjp> NP
14:55 < bubulle> 2.4 2nd stage rebooting....
14:55 < fjp> bubulle: be aware of #290569 for qemu
14:55 < eugen> fjp: it load keymap and then dumps it to /etc/.... Sometimes it dumps incorrect keymap.
14:56 < eugen> fjp: actually install-keymap does it
14:56 < bubulle> ...and 2.4 2nd stage works. That's why this was missed
14:57 < bubulle> <CENSORED>
14:57 < fjp> Oui, c'est vrai.
14:59 * fjp wonders how keymap support in 2.6 is different from 2.4
15:00 < fjp> bubulle: Anything in the d-i logs from kbd-chooser's prebaseconfig?