[09:11] <fish_> hi
[09:12] <fish_> having some trouble with our ubuntu installer.. it fails when running apt-install to add a dkms module. in general it seems like however I call apt-install, it always return with code 1 and prints a bunch of whitespaces
[09:25] <fish_> phew, finally got strace on that box..
[09:28] <fish_> looks like it's reading bunch of cdebconf/ files, like questions.dat
[09:29] <fish_> then some "/etc/newt/palette" related things.. sounds like a terminal control issue
[09:34] <cjwatson> Can you post the entire script that calls apt-install here?
[09:34] <cjwatson> Terminal control is a red herring.
[09:35] <fish_> iit's just apt-install pm80xx-dkms
[09:35] <cjwatson> (An understandable one, but still.)  That's just the effect of apt-install using the debconf confmodule.
[09:35] <cjwatson> I need more context.
[09:35] <cjwatson> Like, where is that script being run?
[09:35] <fish_> same on a interactive shell
[09:35] <fish_> it's a custom partitioner
[09:36] <fish_> but how might that be related? it looks like apt-install always fails with 1 and those empty lines at stdout
[09:36] <cjwatson> But that's not a failure.
[09:36] <cjwatson> (Well, except for the exit code, but that's arguable either way.)
[09:37] <fish_> it's expected to print those empty lines?
[09:37] <cjwatson> Yes.
[09:37] <cjwatson> The debconf confmodule clears the screen on startup.
[09:37] <cjwatson> Anyway, red herring!
[09:37] <fish_> ah got it
[09:37] <cjwatson> You're in the partitioner, and apt-install's purpose is to install packages into the target system.  In a partitioner, you don't *have* a target system yet, so apt-install instead queues the package for later installation.
[09:38] <cjwatson> And it exits 1 to indicate (a bit crudely) that the package isn't ready for use yet.
[09:38] <cjwatson> You should see that it has appended the package name to /var/lib/apt-install/queue.
[09:38] <fish_> yeah.. I have to say it's not my partitioner but supposently it was working before :)
[09:38] <fish_> yep
[09:38] <cjwatson> You probably just need to do "apt-install pm80xx-dkms || true".
[09:38] <fish_> that's true
[09:39] <cjwatson> Maybe something started checking exit codes a bit harder.
[09:39] <fish_> hrmm yeah possibly.. but would be interested to know why it's returning 1
[09:39] <cjwatson> Well, it's right there in the shell script.
[09:39] <cjwatson>         exit 1 # Return error as the package is not ready to be used yet.
[09:39] <fish_> oh
[09:39] <fish_> well yeah.. that explains it.. interesting breakage though.. pretty sure others ran into this as well
[09:40] <cjwatson> Now, it could be argued either way, but it's pretty established; that code dates back to 2004.
[09:40] <fish_> but yes, that happened first time after I upgraded the installer to fix another issue (segfauling wget)
[09:41] <fish_> but as you said, some recent change makes something more strict when it comes to checking return codes
[09:41] <cjwatson> Not sure what that would have been, but anyway, apt-install's exit status has always needed to be ignored when run before the target system has an /etc/apt/sources.list.
[09:41] <cjwatson> If it worked before it was luck :)
[09:42] <fish_> it definitely worked before
[09:42] <cjwatson> Depends how the custom partitioner was hooked in, I guess.
[09:42] <fish_> but since my former co-worker just added the kernel and initrd as blob to our repo, I have no clue which version it used before ;)
[09:48] <fish_> *sigh* such a NIH syndrom project.. this 'custom partitioner' has even it's own partitioning configuration file, similar to partman just a tiny bit easier to understand
[09:49] <fish_> cjwatson: but you where totally right. looks like this error checking was introduced on our side
[09:49] <cjwatson> Aha!
[09:49] <cjwatson> That would explain it. :-)
[09:50] <cjwatson> I couldn't think of anything other than main-menu that could possibly have been relevant on our side, and it's hardly changed at all in years.
[09:54] <fish_> yeah, it explicitly dies.. just a coincidence that I upgraded the installer yesterday and today saw it the first time.. but we probably haven't reinstalled systems since when this was introduced
[09:55] <fish_> which shows nicely why you shouldn't introduce things without testing them..
[12:21] <blut> I am looking at the https://wiki.ubuntu.com/Installer/Development/ToDo right now
[12:22] <blut> funny, that I just modified the initrd to set up my network and to run a custom partitioning
[12:22] <blut> I was unable to make kickstart and d-i do the work
[12:23] <blut> for the network e.g. I need a dhcp ip and from that one I determine a second IP and only then set up the gateway
[12:23] <blut> I will write a documentation on how to do that
[16:29] <flexiondotorg> cyphermox, Could you take a look at a merge proposal for me please - https://code.launchpad.net/~ubuntu-mate-dev/ubiquity/mate-compatibility/+merge/297180
[16:30] <flexiondotorg> You might remember that Ubuntu MATE has not been following recommends. Well, I've unpicked that but there is some fallout ;-)
[16:32] <flexiondotorg> cyphermox, And this https://code.launchpad.net/~ubuntu-mate-dev/livecd-rootfs/mate-compatbility/+merge/297219
[16:32] <cyphermox> yeah I saw the changes for flipping that
[16:32] <flexiondotorg> cyphermox, And also this https://code.launchpad.net/~ubuntu-mate-dev/debian-cd/mate-compatbility/+merge/297222
[16:33] <cyphermox> hrm, firefox is slow when there's so much stuff running at once :)
[16:33] <flexiondotorg> Because cloudtop is not required now.
[16:33] <flexiondotorg> Chromium.
[16:33] <cyphermox> cloudtop?
[16:33] <cyphermox> oh, I see yeah ok
[16:34] <cyphermox> are you sure for ltsp you don't want to keep cloudtop?
[16:34] <flexiondotorg> cloudtop was so similar to desktop.
[16:35] <flexiondotorg> And desktop and now much smaller.
[16:35] <flexiondotorg> It was for an integrator and they've requested the change.
[16:35] <flexiondotorg> So desktop and cloutop (their term) installs are consistent.
[16:35] <cyphermox> the ubiquity change looks off, it should already have been working
[16:36] <cyphermox> (what I mean by that is this leads to questioning why u-s-d gets pulled in to your images)
[16:36] <flexiondotorg> cyphermox, Because authentication plugins eventually require u-s-d.
[16:37] <cyphermox> well, they should probably grow an alternate depends instead
[16:37] <flexiondotorg> And having u-s-d on the image is fine.
[16:37] <flexiondotorg> I just don't want it to run in ubiquity-dm for the MATE images, because the desktop/font settings are all weird.
[16:37] <cyphermox> well, if it's unnecessary and doesn't run...
[16:38] <flexiondotorg> u-s-d does run.
[16:38] <cyphermox> ah?
[16:38] <flexiondotorg> But I don't want it to.
[16:38] <cyphermox> right, that's what I mean
[16:38] <cyphermox> if it's not to be running on the image, why include it?
[16:39] <cyphermox> ie. the fact that there is a change needed in ubiquity points to that there may be an issue elsewhere.
[16:39] <flexiondotorg> cyphermox, Because the authentication plugins require it. And there is no alternative to satisfy that dependency.
[16:40] <cyphermox> what are these authentication plugins?
[16:42] <flexiondotorg> libaccount-plugin-1.0-0
[16:43] <flexiondotorg> Which Recommends: Recommends: unity-control-center-signon
[16:43] <flexiondotorg> Which Depends: unity-control-center
[16:44] <flexiondotorg> I have temporarily removed the packages (plugins for shotwell) from the seeds and meta packages to prevent this happening.
[16:44] <cyphermox> ok
[16:45] <flexiondotorg> And I also have this merge proposal - https://code.launchpad.net/~ubuntu-mate-dev/indicator-session/mate-compatibility/+merge/297183
[16:45] <cyphermox> what pulls in libaccount-plugin-1.0-0?
[16:46] <flexiondotorg> account-plugin-facebook -> libaccount-plugin-generic-oauth -> libaccount-plugin-1.0-0
[16:46] <flexiondotorg> And other account-plugin-* packages do the same.
[16:47] <cyphermox> and what pulls in account-plugin-facebook?
[16:48] <flexiondotorg> Shotwell suggests it.
[16:48] <flexiondotorg> And if included allow photo to be uploaded to your Facebook account.
[16:48] <cyphermox> what I'm getting at here is that the who dependency chain looks odd -- all these account-plugin-* are Ubuntu Online Account things that probably don't belong at all (and won't work) if you don't install u-s-d
[16:49] <flexiondotorg> Which I'd like to include because, as far as my wife is concerned, Internet ~= Facebook
[16:49] <flexiondotorg> Right, I do want u-s-d installed. Because as you say without it, those plugins don't work.
[16:49] <cyphermox> well you don't really need to care about things shotwell recommends
[16:49] <flexiondotorg> There is no UI for Shotwell to authenticate with "whatever"
[16:49] <cyphermox> flexiondotorg: if it's not *running*, I'm doubtful they will work any better.
[16:50] <flexiondotorg> By seeding account-plugin-facebook, u-s-d is on the live image.
[16:50] <cyphermox> yes, I know that
[16:50] <flexiondotorg> And when ubiquity-dm starts, it prefers u-s-d.
[16:50] <cyphermox> it doesn't matter
[16:51] <cyphermox> if u-s-d isn't running on the installed system, I don't think you'll be able to login to facebook to push photos from shotwell
[16:52] <cyphermox> but it looks like sufficient justification for the ubiquity change, I'm not here to dictate what you can or can't or shouldn't have on your image, just to plant the seed of doubt and make sure a ubiquity change isn't pure crack ;)
[16:54] <cyphermox> I will leave debian-cd to infinity to review, or someone else on the ubuntu-cdimage team.
[16:54] <flexiondotorg> cyphermox, OK, thanks.