[07:58] <xnox> kentb-out: hm? driver-injection-disk/load has default as "true", so if the driver-injection-disk-detect udeb is present it should auto detect & use OEMDRV disk.
[11:13] <cjwatson> xnox: I'm working on bug 1197766, BTW - I think I see what you did wrong
[11:13] <ubot2`> Launchpad bug 1197766 in OEM Priority Project "Different partition layout after recovery with keep home partition" [Critical,Confirmed] https://launchpad.net/bugs/1197766
[11:14] <cjwatson> need to do the filtering in choose_recipe instead.  Trying a similar-but-different patch
[11:14] <xnox> hm. ok.
[11:15] <xnox> cjwatson: note that at the moment the workaround is to use "0 0 0" for the UEFI partition in the recipe. Then it's reused without affecting calculated size required.
[11:15] <cjwatson> min_size is too low-level to go modifying $scheme
[11:15] <cjwatson> mm, I guess that would work too
[13:19] <kentb> xnox: right. That's what I interpreted as well, but, it seems that if we don't do something like "anna/choose_modules=driver-injection-disk-detect" at the installer command line the dirver disk on USB stick won't get picked up (at least from what I can tell).  I'll be glad to try again without the command line parameters to be sure, but, we won't see
[13:19] <kentb> anything in /var/cache/firmware/ if we don't.  I could completely be missing something, though.
[13:21] <xnox> kentb: interesting, that indicates that driver-injection-disk-detect udeb is missing by default and needs a manual nudge to be fetched. What media are you installing from?
[13:22] <kentb> xnox: usb for the OS and then the driver disk on a separate USB stick, formatted fat32 with an OEMDRV label.  The anna portion of the command line is so far the nudge that it needed to load the driver-injection-disk udeb.
[13:22] <xnox> kentb: "usb for the OS" netinstall, server, mini image?
[13:23] <kentb> xnox: so, if there's a way to get that udeb loaded up by default, that'll be a big help (I think).
[13:23] <kentb> xnox: the server image in this case
[13:30] <xnox> cjwatson: infinity: driver-injection-disk-detect is currently "Priority: optional", is ok to bump it to "Priority: standard" such that it gets auto-loaded by default, as per ^
[13:32] <cjwatson> That's probably fine, but its value in the package makes no difference - it has to be overridden in the archive.
[13:33] <cjwatson> It'll slow down the installer a bit by iterating over all USB partitions.  Probably not significantly.
[13:34] <cjwatson> I dunno.  It does mean that if you leave a malicious stick inserted unwarily, it can silently own your installation, since driver-injection-disk/load is true by default and not displayed.
[13:34] <cjwatson> So it does make me a bit nervous.
[13:41] <xnox> and there is no signature checks / validation of the installed debs on the sticks.
[13:42] <xnox> =/ essentially an open backdoor into otherwise verified installation.
[13:42] <xnox> kentb: ^^^
[13:42] <kentb> xnox: right. understood.
[13:45] <cjwatson> Is it so bad to need to pass anna/choose_modules=driver-injection-disk-detect?  We could add an alias for that if it's just about length.
[13:55] <infinity> It's not wildly intuitive for the "average user".
[13:55] <infinity> But I'd expect a debconf template asking me if I want to use driver disks before scanning and copying willy-nilly.
[13:55] <infinity> (Which then adds another question if we make it a default component, which sucks)
[13:56] <kentb> cjwatson: what infinity said.  Plus we're looks to streamline the process a bit.  RHEL has it built into anaconda to check for those OEMDRV disks automatically.  Big picture, we're looking to get Ubuntu listed in the lifecycle controller on poweredge as an option to install the OS on poweredege servers.  RHEL, Windows, and SLES are on there now.  When you
[13:56] <kentb> select the OS, the lifecycle controller will provide out-of-box drivers via a virtual disk labeled OEMDRV.  It might be possible for the lifecycle controller folks to add that command line automatically, but, we're still working on getting our foot in the door with them so to speak.
[13:57] <infinity> Could we just scan and *if* we find OEMDRV disks, then trigger a yes/no template before doing Very Bad Things?
[13:58] <kentb> infinity: that'd be good, too.  That's how RHEL does it now.  It stops and asks even with the OS selected in lifecycle controller.
[14:04] <cjwatson> Yeah, if there were UI then I'd be less unhappy
[14:04] <cjwatson> (Preseedable, of course)
[14:05] <cjwatson> That's probably just a matter of lifting the priority of that question to high.
[14:05] <cjwatson> It's medium right now.
[14:06] <kentb> I'm game for whatever you guys think is the safest way to do it
[14:12] <infinity> cjwatson: Well, raising it to high would trip it all the time, wouldn't it?
[14:13] <infinity> cjwatson: (I haven't looked, but I assume it asks before scanning currently, rather than after)
[14:13] <cjwatson> No, it's only displayed if there's actually something to do
[14:13] <cjwatson>         if [ "$label" = "OEMDRV" ]; then
[14:13] <infinity> Ahh.
[14:13] <cjwatson>                 db_input medium driver-injection-disk/load || true
[14:13] <infinity> Well, that would be perfect, then.
[14:20] <kentb> indeed!
[14:24] <xnox> cjwatson: raise the priority on the question to high, raise the priority on the package, keep the default at true?
[14:26] <cjwatson> Sounds right
[14:36] <xnox> ack, will update merge proposal.
[18:57] <rostam> HI all,  so I have created a list of installed packages on my system: "dpkg --get-selections > my-packages". Now I have created kickstart/preseed iso image installing ubuntu 12.04 LTS.  How could I install packages listed in 'my-packages' during kickstart installation? thx
[21:13] <rostam> Hello, during custom installation Ubuntu 12.04 LTS server, I still get asked: "Write the changes to disks?" how could I preseed that please? thx
[21:32] <maxb> d-i partman/confirm boolean true
[21:32] <maxb> d-i partman/confirm_nooverwrite boolean true
[21:32] <maxb> rostam: ^
[21:33] <maxb> You can include just the _nooverwrite line to skip prompting only if there are no existing partitions that will be overwritten (fresh empty drives)
[22:00] <rostam> maxb: I did have the first d-i, but not the second one so I will try that now. Thanks
[23:32] <jbicha> xnox: ubiquity's autopkgtest fails because python3-oauthlib isn't installed, you can add dependencies to debian/tests/control if you need to
[23:32] <jbicha> https://bazaar.launchpad.net/~ubuntu-installer/ubiquity/trunk/revision/5989