=== psivaa_afk is now known as psivaa === davmor2_ is now known as davmor2 [09:57] I wanted to use a server CD as a substitute for the alternate CD which is no longer produced, but despite preseeding tasksel/first in the same way the desktop CD does, I didn't end up with a desktop being booted into [09:57] Any suggestions on what the missing piece is? [09:58] Server does some slightly odd things; its preseeding of tasksel/force-tasks might be relevant. [09:58] Also, our CDs tend to only permit installing things from the CD itself. [09:58] You might be better off using the netboot mini.iso. [09:59] Since you're going to end up downloading a bunch of stuff anyway if you're installing desktop from a server image. [10:00] Hm. I did consider mini.iso, but I wanted the benefit of a grub setup to make my media EFI-bootable too [10:01] The mini.isos should be hybrid, I thought ... [10:01] The amd64 one, anyway [10:01] Also, I thought I'd managed to remove the server preseed stuff by substituting my own preseed file and changing the preseed file= boot option in the image [10:01] debian-installer (20100211ubuntu22) maverick; urgency=low [10:01] ... [10:01] * Add GRUB EFI configuration on amd64, and make mini.iso images there [10:01] multi-catalog. [10:01] -- Colin Watson Fri, 27 Aug 2010 10:11:36 +0100 [10:02] Yeah, it's probably more the restriction to what's on the CD that's relevant. [10:02] Ah, whoops. I assumed the mini.iso was syslinux-only by definition [10:03] The server-based install I attempted definitely installed lots of desktop packages, it just didn't boot into X [10:05] Baffling, "tasksel tasksel/first multiselect ubuntu-desktop" somehow managed to get unity (amongst other things) installed but not ubuntu-desktop itself [10:06] It'll be installing whichever bits of the task it can find, I suppose. Perhaps "d-i pkgsel/include string ubuntu-desktop" would do better [10:06] That'll go for the metapackage [10:08] Worth a try, though I'm intrigued how it managed to the subset it did [10:09] Seems odd indeed. I could look at a syslog ... [11:00] It looks like it really did only install the bits of the task that were on the CD [11:01] But then the execution of /usr/lib/finish-install.d/01oem-config-udeb pulled assorted desktopish bits in [11:01] That makes sense, since it was looking for the task header in its known packages, which was only from the CD. [11:01] Right. I think you will find it very hard to make this work right from the server CD. [11:01] mini.iso or a netinst would definitely be a saner way to install desktops via d-i. [11:01] * maxb will change course and play with mini.iso [11:02] s/netinst/netboot/ [11:02] This is unfortunately basically a case of galloping requirements clashes back in the early days of Ubuntu. [11:02] I might do it differently nowadays ... [11:24] Hmm. The mini.iso *does* include an EFI grub, but it doesn't include the /EFI/BOOT/BOOTx64.EFI and /EFI/BOOT/grubx64.efi files that the other boot media have [11:26] I think it's laid out a bit differently. It should still boot ... [11:31] It should boot in BIOS mode, but I can't see how it's going to work in EFI mode [11:54] I thought it had the relevant bits embedded in an El Torito catalog entry. [11:54] It's been a while. [12:20] Ah [12:20] It is entirely possible, but that wouldn't survive the translation from .iso to USB stick === kentb-out is now known as kentb === tjaalton_ is now known as tjaalton === cr3_ is now known as cr3 === cr3 is now known as Guest77844 === kentb is now known as kentb-out === cjwatson_ is now known as cjwatson