Merge lp://qastaging/~abone/ubuntu/precise/console-setup/fix-433897 into lp://qastaging/ubuntu/precise/console-setup

looks good. There is a typo in the comment 'If /etc/default/grub has "spalsh" option in' s/spalsh/splash/.
Will test and merge if this fixes the problems.

I'd like to have some time to review this before it lands, as this is a very delicate area. I'll try to make time for it next week.

i dont think it is the right approach to forcefully set FRAMEBUFFER=y as soon as splash is set on the commandline, imho the setting of the font should not depend on the type of output ...

also note that FRAMEBUFFER=y installs additional plymouth theme bits in the initrd growing its size by quite a bit.

the better option would be to simply not depend on FRAMEBUFFER for the console-setup hook (if thats doable without breakage) and install it regardless

I agree that it is better if setting up font does not depend on FRAMEBUFFER, but don't know if OPTION may be safely removed. My knowledge of linux console is rather weak. Will try to test.

I've tried to remove OPTION=FRAMEBUFFER from init-top/console_setup script. At least it does not break my system. This option was added by Scott James Remnant <email address hidden> in revision 69 on 15th December 2009 with comment "We don't need the initramfs hooks if the initramfs doesn't load the framebuffer or splash screen" Looks like a kind of optimization.

So, I don't mind having console_setup enabled all the time. Another way, if you prefer to save some space in latin locales, is to creaate separate option for console-setup, and enable it regardles of framebuffer usage. For example it can be enabled if /etc/default/locale contains some non-latin locale or /etc/default/console-setup have some "non-default" font setup.

My patch also contains upstart job for console-setup. This job, in my opinion, should be enabled anyway.

Waiting for your feedback.

After some thought, I agree that this Upstart job is the right thing to add, and I'm going to apply this to quantal. Thanks!

I am much less convinced by the proposal to remove OPTION=FRAMEBUFFER (and still less by the proposal to force FRAMEBUFFER=y when splash is enabled; I agree with Oliver on that). I also don't think that creating a separate option for it is the right answer. If the framebuffer isn't enabled for some other reason, then the console-setup script should not need to run in the initramfs at all (except in the case of a panic, which is handled); the fact that this shouldn't be needed is what OPTION=FRAMEBUFFER expresses. It should be sufficient for the corresponding code to run from udev rules and Upstart jobs.

Therefore, if the udev rules and Upstart jobs aren't working properly when there is no framebuffer code in the initramfs, that is what needs to be debugged. I believe this ought to be possible without running console-setup code in the initramfs.

I have merged the Upstart job part of this into lp:~ubuntu-core-dev/console-setup/ubuntu, which is the development branch we use. For future console-setup work, please branch from and propose to there.

I think it's least confusing (although no option avoids confusion altogether) if I mark this merge proposal as merged at this point, and suggest that you file something separately with whatever arises from my comments about OPTION=FRAMEBUFFER. These seem to be separate issues to some extent, so let's treat them separately.

Ok, I've tested console-setup version 1.70ubuntu6. I think it will solve the bug 433897 for most users. There is some small issue with mount all and fsck on boot (both print translated messages while font it configured), but I think most people can live with it.

Thank you.