On Sat, Jan 17, 2004 at 03:56:01PM +0100, Bastian Blank wrote: > On Fri, Jan 16, 2004 at 07:13:41PM +0100, Gaudenz Steinlin wrote: > > Another possible solution which would be less intrusive to core > > components like anna would be to handle this via the isinstallable > > scripts in the quik-installer package. yaboot-installer currently does > > this to make sure it's only used on newworld powermac.
Gaudenz, i have to shatter your illusions here, yaboot-installer does nothing such. Now that the new -5 powerpc kernels have been used in linux-kernel-di, i decided to give the installation a try on my pegasos box, and here is what happened : I easily built the initrd enabled kernel using the netboot initrd (the cdrom one is too big). I booted it with the devfs=mount option, and it worked well. First problem appeared as i got to choose the keyboard. Only USB-pmac keyboard were available, while i need ps2 keyboards. This resulted in me choosing the USB-pmac french keyboard, which worked some, but either Alt or the function keys were not mapped, so i couldn't switch to the second console. Which gives : Problem 1 : keyboard selection ========= => Easy solution. Propose both ps2 and USB-pmac (and maybe amiga for apus) keyboard as choice for powerpc. I think the oldworld adb keyboards get emulated as USB-pmac ones, so this should be no problem. Someone with oldworld hardware would have to confirm though. Then everything went well, there was some issue with my sym53c8xx scsi card not being detected, so i could not test the partitioning scheme on my scsi removable disk, but i guess that this should be ok, since i can use parted. Problem 2 : partitioning ========= => I don't know if parted is used on all powerpc subarches, but i guess it could at least be possible. I called parted by hand from a shell (and thus discovered that Alt+F2 did not work) and it worked well. That said, maybe there is some message or something which should tell the user what kind of partition table should be chosen, depending on subarches. So, i chose a partition of my main disk, taking the risk of again losing everything, and proceeded with the install. This went well, upto the point where yaboot-installer was to be run, and naturally complained about no apple bootstrap partition or something. Problem 3 : bootloader ========= => Here the bootloaders, quik for oldpmac, yaboot for newpmac, nothing for the other subarches, need to check for the subarch. This can be done by either the isinstallable script, or the libd-i arch detection code. I guess that the best would be the creation of smallish udebs for the other subarches without bootloader, which could simply tell the user that there is no bootloader for its subarch and the steps needed to boot the final kernel. This could include information on where the kernel has been installed and such. Ok, this is mostly what needs to be done to enable support for chrp, prep, chrp-rs6k and chrp-pegasos powerpc subarches. Not really all that much. I rebooted the box by chosing "finish and reboot", altough the french translation let me a bit dubious about if it was the right thing to do, but i will tell that the french translation team once it is done, and then proceeded to reboot into the partition by hand. And there i faced the last problem, well most probably a bug or something, i was simply presented with a login prompt, and i had not the root password to enter it. Normally at this time, it should have booted in the post-d-i install process, and proceeded smoothly with the install. Why this happened has me trully baffled, but i couldn't do much, for lack of root password. > > As these little udebs are really small it does not really matter much if > > the get installed by anna or not. No, as long as they don't get used when they should not. > Sven Luther and myself decided to enable the subarchitecture matching in > anna and libdi. The yaboot-installer changes should propagate into > unstable tomorrow, the other changes may need some days. Cool. Friendly, Sven Luther -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
