[00:41] <cjwatson> slangasek: grep -B1 -A2 db_text /usr/share/debconf/confmodule
[00:58] <doko> now afk, getting some food, will be back in an hour or so
[01:03] <slangasek> cjwatson: heh, ok
[01:48] <CIA-32> ubiquity: stgraber * r5290 ubiquity/ (debian/changelog ubiquity/frontend/gtk_ui.py): Only allow for one orca instance at a time.
[02:20] <infinity> stgraber: How that you're a GTK3 CSS expert, want to port Human to GTK3 for me?
[02:20] <infinity> stgraber: Thanks in advance.
[08:53] <CIA-32> ubiquity: cjwatson * r5291 trunk/ (3 files in 2 dirs):
[08:53] <CIA-32> ubiquity: Make sure to load translations for "USB" and "CD" on the language page
[08:53] <CIA-32> ubiquity: (LP: #958249).
[12:52] <ev> dear lord. Do you people sleep?
[12:53] <stgraber> :)
[12:56] <ev> so as it turns out, switching on crash reporting for whoopsie has an effect: https://bugs.launchpad.net/ubuntu/+source/whoopsie-daisy
[12:58] <bdmurray> heh
[13:01] <jodh> lo
[13:06] <cjwatson> stgraber: bug 961003 looks like a regression from your keymap indicator work
[13:06] <ubot2`> Launchpad bug 961003 in ubiquity "ubiquity crashed with TypeError in restrict_list(): 'set' object has no attribute '__getitem__'" [High,New] https://launchpad.net/bugs/961003
[13:07]  * cjwatson takes bug 961025
[13:07] <ubot2`> Launchpad bug 961025 in ubiquity "ubi-webcam: Pictures listed on 1 column" [Medium,Triaged] https://launchpad.net/bugs/961025
[13:07] <stgraber> cjwatson: looking
[13:11] <stgraber> cjwatson: argh, weird we never hit that case before... I'll add a test and fix it
[13:12] <cjwatson> I wondered if it was due to successively choosing different languages
[13:13] <cjwatson> I'm attacking 961025 to start with by way of a little test suite mod (http://paste.ubuntu.com/893578/) that causes it to show me all the pages in sequence
[13:13] <cjwatson> with any luck that will help
[13:14] <cjwatson> actually, what would people think about committing a cleaned-up version (http://paste.ubuntu.com/893582/)?
[13:14] <stgraber> no, it's just that it's the only part where I'm trying to get a subset of a set instead of a list. It's in the code that triggers when default-layout+country-layouts+simple-variants still makes less than 4 entries, it then adds the next four ones from the original list
[13:15] <cjwatson> right, I was just wondering if the keymap list grew over time or something
[13:15] <cjwatson> certainly the actual fix is probably trivial
[13:16] <bdmurray> the 'install ubuntu 12.04 lts' icon text has gotten longer? I now see that an 'install this system permanently to your hard disk'
[13:17] <cjwatson> heh, not that it wants to show me the webcam page anyway, grumble
[13:17] <jibel> bdmurray, bug 961141
[13:17] <ubot2`> Launchpad bug 961141 in nautilus "Desktop icons display comments from desktop file in label" [Undecided,New] https://launchpad.net/bugs/961141
[13:17] <jibel> known upstream but I couldn't find an existing bug report.
[13:18] <stgraber> cjwatson: yeah, adding an option to make the test slower sounds like a good idea when we're debugging UI weirdness (like our usual won't-fit-on-netbook-screen)
[13:19] <bdmurray> jibel: thanks
[13:24] <CIA-32> ubiquity: cjwatson * r5292 trunk/tests/test_gtkui.py: make it easier to debug layout problems using 'UBIQUITY_TEST_SHOW_ALL_PAGES=1 tests/run --no-xvfb test_gtkui'
[13:25] <CIA-32> ubiquity: cjwatson * r5293 trunk/tests/test_gtkui.py: the webcam page seems to need two ui.refresh() calls before it appears here; who knows why, but this is fairly harmless
[13:25] <cjwatson> of course this still shows webcam images in two columns *grumble*
[13:28]  * ogra_ wishes we had a proper way to actually check for a webcam
[13:28] <ogra_> but i couldnt find something better than the code already has
[13:29] <ogra_> (for bug 924419)
[13:29] <ubot2`> Launchpad bug 924419 in ubiquity "oem-config detecting camera where no camera exists" [Medium,Incomplete] https://launchpad.net/bugs/924419
[13:29] <cjwatson> checking for a capture device doesn't seem awful?
[13:29] <cjwatson> ah, I see :-/
[13:29] <ogra_> well, if the driver exposes a capture device :)
[13:29] <cjwatson> does cheese do anything better?
[13:30] <cjwatson> and is it cheap to check whether the device works?
[13:30] <ogra_> but there doesnt seem to be something like a check for "hey you got something wired up there"
[13:30] <stgraber> I really think the bug is on the kernel's side... I'm sure having these /dev/videoX devices will confuse a lot more than just ubiquity
[13:30] <ogra_> cjwatson, i wouldnt know how apart from take an actual pic and check if there is something on it but black pixels
[13:32] <ogra_> essentially the bug is on the HW side (or the driver as stgraber said), it shouldnt expose a device unless there is actually something wired up
[13:44] <CIA-32> ubiquity: cjwatson * r5294 trunk/tests/run: add a --no-build option to skip building source code first
[13:48] <cjwatson> ev: is the order of existing images in the face selector deliberately pseudo-random (filesystem order)?  I was thinking about sorting it so that it would be more reliably testable - http://paste.ubuntu.com/893625/
[13:48] <CIA-32> ubiquity: stgraber * r5295 ubiquity/ (debian/changelog tests/test_misc.py ubiquity/misc.py): set_indicator_keymaps: It's not possible to access a position in a set so convert it back to a list.
[13:49] <cjwatson> maybe s/pseudo-random/arbitrary/
[13:49] <ev> cjwatson: it wasn't purposeful, no. So by all means
[13:49] <cjwatson> OK, cool
[13:50] <cjwatson> not that I've figured out where jibel's bug there is coming from yet - just noticed that in his screenshot the astronaut image was first, whereas for me it was lightning
[13:51] <doko> good morning
[13:52] <CIA-32> ubiquity: cjwatson * r5296 trunk/ (debian/changelog ubiquity/gtkwidgets.py): Sort existing images in the face selector by filename.
[13:54]  * stgraber takes bug 960628
[13:54] <ubot2`> Launchpad bug 960628 in ubiquity "guided partitioner allows a partition size smaller than minimum requirement" [Undecided,Confirmed] https://launchpad.net/bugs/960628
[13:55]  * barry reboots mumble
[13:56] <cjwatson> does -usbdevice work for anyone else to attach a webcam to a kvm?
[13:56] <cjwatson> I have 'sudo kvm -monitor stdio -m 512 -cdrom precise-desktop-i386.iso -hda t.img -usb -usbdevice host:18ec:3299' but I get 'kvm: usb_create: no bus specified, using "usb.0" for "usb-host"' and 3x 'kvm: Warning: speed mismatch trying to attach usb device USB2.0 PC CAMERA to bus usb.0', and the virtual machine can't see the webcam
[13:58] <cjwatson> I guess maybe I can stub it out somehow
[13:59] <bdmurray> I find it really strange that if I screw up ubi-partman.py the page isn't displayed at all and ubiquity carries on
[13:59] <stgraber> cjwatson: most webcams are usb2 only so won't work with kvm's usb1 hub. There's a way of adding a usb2 hub to qemu but last I tried it was very experimental and crashed kvm for me
[13:59] <cjwatson> stgraber: gotcha, thanks
[13:59] <cjwatson> ubiquity's plugin interface isn't perhaps the most robust ...
[14:00] <stgraber> if anyone has an idea on what might be the source of bug 950282 any help would be appreciated
[14:00] <ubot2`> Launchpad bug 950282 in oem-priority/precise "Installation failing with pop-up "The installer encountered an unrecoverable error and will now reboot."" [Critical,Confirmed] https://launchpad.net/bugs/950282
[14:01] <stgraber> so far it seems to be pretty random, shows up as a libc segfault and the source is still pretty much unknown
[14:01] <stgraber> jibel did some tests but without being able to track down the change introducing the bug (that I'm aware of)
[14:06] <cjwatson> bah, 961025 shows up in kvm with the webcam availability check stubbed out, but not in the test suite
[14:07] <CIA-32> ubiquity: stgraber * r5297 ubiquity/debian/changelog: Add bug number for panel bug
[14:08] <cjwatson> I suppose it might have something to do with the presence of the progress bar below
[14:12] <cjwatson> yes!
[14:12] <jibel> stgraber, I can't reproduce it since the 14th on the same hardware that was affected (tested on 2 machines, same brand, same model)
[14:15] <stgraber> doko: can you take a quick look at bug 950282? when it occurs we get a libc segfault in the syslog, sadly it's kind of hard to reproduce as even jibel can't get it to happen (but OEM appears to be able to reproduce fairly easily)
[14:15] <ubot2`> Launchpad bug 950282 in oem-priority/precise "Installation failing with pop-up "The installer encountered an unrecoverable error and will now reboot."" [Critical,Confirmed] https://launchpad.net/bugs/950282
[14:16] <cjwatson> in fact that's completely bizarre.  allow_change_step(False) is enough to change the layout
[14:18] <CIA-32> ubiquity: cjwatson * r5298 trunk/tests/test_gtkui.py: test_pages_fit_on_a_netbook: disable changing step after the partman page
[14:30] <CIA-32> ubiquity: stgraber * r5299 ubiquity/ (debian/changelog ubiquity/frontend/gtk_ui.py): Call show_all() instead of show() when starting the quit dialog to fix bug 960600
[14:30] <ubot2`> Launchpad bug 960600 in ubiquity "Quit dialog - broken geometry" [Undecided,Fix committed] https://launchpad.net/bugs/960600
[14:31] <cjwatson> nice catch
[14:33] <doko> stgraber, from the bug log it looks like it's not reproducible anymore?
[14:34] <doko> jodh, about bug #946663, I'm now able to reproduce it here locally
[14:34] <ubot2`> Launchpad bug 946663 in ubiquity "Installer stuck at "Removing conflicting operating system files..."" [High,Confirmed] https://launchpad.net/bugs/946663
[14:35] <stgraber> doko: according to comment #12 and #14, it still happens for Christopher (OEM) on DELL machines
[14:39] <jodh> doko: yes, I saw your comment. I'm currently looking at bug 684309 whilst waiting to install a kvm image with an encrypted /home.
[14:39] <ubot2`> Launchpad bug 684309 in ubiquity "Ubiquity Advanced Partitioning - Converting/Using ext4 to ext2 without formating formats the partition without warning" [Undecided,Confirmed] https://launchpad.net/bugs/684309
[14:41]  * ogra_ rolls an ubuntu-server arm image since last nights builds failed due to ubiquity being building 
[14:42] <bdmurray> has something changed that would have fixed bug 960278?
[14:42] <ubot2`> Launchpad bug 960278 in ubiquity "ubiquity crashed with UnicodeEncodeError in debug(): 'ascii' codec can't encode character u'\xe9' in position 55: ordinal not in range(128)" [Undecided,New] https://launchpad.net/bugs/960278
[14:43] <slangasek> bdmurray: are you running your test in French?
[14:43] <bdmurray> oui
[14:44] <slangasek> hmm, ok
[14:44] <cjwatson> possibly the incorporation of the new console-setup upload
[14:44] <cjwatson> which was in 2.9.31
[14:44] <cjwatson> exactly how that would have fixed that, not entirely certain
[14:48]  * slangasek wonders if we should throw some random ascii look-alike characters into our English strings in the installer to exercise these code paths earlier in development :)
[14:48] <bdmurray> well how would I find an Unknown keyboard variant for a layout?
[14:48] <cjwatson> slangasek: python3 should address it, when we port
[14:49] <slangasek> yeah
[14:49] <cjwatson> its number one benefit AFAIAC
[14:49] <cjwatson> (with a fairly significant cost, but still ...)
[14:50]  * cjwatson attempts to port the face selector to GtkGrid in the hope that height-for-width geometry management will fix 961025
[14:53] <bdmurray> okay, so I got it to crash again now
[15:01] <doko> jodh, need to reboot, is there more information to provide?
[15:02] <jodh> doko: I don't think so at this stage. It's looking ecryptfs-related as I haven't seen it with non-encrypted /home yet.
[15:05] <doko> mumble eating 1.5gb ram ...
[15:06] <slangasek> doesn't do that here
[15:06] <slangasek> oh, well, depending on how you're measuring I guess
[15:06] <slangasek> 1395m VIRT
[15:09] <ogra_> well, its a virtual sprint, he likely needs all his virtual mem ;)
[15:28] <stgraber> cjwatson: if you have a minute, can you review http://paste.ubuntu.com/893749/ ?
[15:28] <cjwatson> Any objections to http://paste.ubuntu.com/893748/ ?  I've spent a good hour trying to make it work more elegantly with proper geometry management and failed; I think this should do for now.
[15:28] <stgraber> cjwatson: basically trying to make the minimum install size consistent and using the same function everywhere
[15:29] <stgraber> cjwatson: looking
[15:30] <stgraber> cjwatson: no objection, seems to be the easiest way to enforce the two columns layout ;)
[15:30] <cjwatson> yeah, ineed
[15:30] <cjwatson> *indeed
[15:30] <cjwatson> stgraber: yours seems fair enough to me
[15:30] <cjwatson> and I second removing code quasi-duplication ...
[15:31] <CIA-32> ubiquity: cjwatson * r5300 trunk/ (debian/changelog ubiquity/gtkwidgets.py):
[15:31] <CIA-32> ubiquity: Force face selector icon view to a two-column layout, in lieu of better
[15:31] <CIA-32> ubiquity: geometry management (LP: #961025).
[15:31] <CIA-32> ubiquity: cjwatson * r5301 trunk/debian/changelog: typo
[15:34] <CIA-32> ubiquity: stgraber * r5302 ubiquity/ (4 files in 3 dirs): Move min_size() from ubi-prepare to install_size() in misc, update any remaining code needing the minimal install size to using it and tweak ubi-partman's resize function to use it too.
[15:36]  * cjwatson holds his nose and dives into bug 684292
[15:36] <ubot2`> Launchpad bug 684292 in ubiquity "Installing to USB drive (sdb), writes MBR to internal drive (sda) without warning or giving option to change" [High,Triaged] https://launchpad.net/bugs/684292
[15:56]  * stgraber starts poking at bug 960039
[15:56] <ubot2`> Launchpad bug 960039 in libindicator "ubiquity greeter : Label of the network indicator changes to the keyboard layout code" [Undecided,Confirmed] https://launchpad.net/bugs/960039
[16:32] <cjwatson> ev: so, I don't think this bug can actually be in misc.grub_default() after all, but are you OK with http://paste.ubuntu.com/893845/ anyway?  Seems a shame to waste the effort writing test code
[16:33] <cjwatson> (It also splits up misc.grub_default a bit to make it more testable, but there should be no functional change)
[16:33]  * ev looks
[16:40] <ev> this is excellent
[16:40] <ev> +1
[16:40] <ev> yay for very readable tests
[16:41] <cjwatson> coolio, thanks
[16:41] <cjwatson> giant test framework plus short tests ;-)
[16:41] <CIA-32> ubiquity: cjwatson * r5303 trunk/ (debian/changelog tests/test_misc.py ubiquity/misc.py):
[16:41] <CIA-32> ubiquity: Refactor ubiquity.misc.grub_default to make it testable, and add tests
[16:41] <CIA-32> ubiquity: for it.
[16:41] <ev> :)
[16:53] <ogra_> cjwatson, so i dont seem to get around the full username, login and tasksel questions in oem-config using the preseed file from the bug
[16:54] <ogra_> https://bugs.launchpad.net/ubuntu/+source/ubiquity/+bug/924018/+attachment/2702031/+files/preseed.cfg
[16:54] <ubot2`> Launchpad bug 924018 in ubiquity "Preseeding doesn't work with oem-config" [Medium,In progress]
[16:55] <ogra_> nothing special in there i think, so i suspect there are still areas in the user-setup and tasksel plugins that dont let it do full automation
[16:56] <cjwatson> ok, do you have a debug log?
[16:57] <cjwatson> (it's dinnertime here now, but I can look right afterwards)
[17:00] <ogra_> just added to the bug
[17:00] <ogra_> funnily its only asking full name and login, not the password question
[17:11] <ogra_> oh, i just notice we dont actually use the normal user-setup but something called user-setup-ask-oem
[17:12] <CIA-32> ubiquity: stgraber * r5304 ubiquity/ (debian/changelog ubiquity/frontend/gtk_ui.py): Switch from calling 'tail' to calling 'busybox tail' so the ubiquity vte widget shows something useful during installing.
[17:14] <slangasek> lunch
[17:16] <barry> so, i think https://code.launchpad.net/~barry/ubiquity/bug-959724/+merge/98515 is ready for review.  see the comment about the tests.  /me -> lunch
[17:18] <stgraber> barry: can you add a changelog entry?
[17:19] <stgraber> I'm going for lunch now too, will have a closer look when I'm back
[17:19]  * jodh -> food.
[17:20] <ogra_> oh, funny, so /usr/lib/ubiquity/user-setup/user-setup-ask and /usr/lib/ubiquity/user-setup/user-setup-ask-oem differ only in the reserved-usernames they read from either /usr/lib/ubiquity/user-setup/reserved-usernames or /usr/lib/ubiquity/user-setup/reserved-usernames-oem
[17:20] <ogra_> thats a huge amount of duplicated code for just a single grep
[17:21]  * bdmurray goes to lunch
[17:44]  * ev looks over barry's merge
[18:05] <cjwatson> barry: fat16/fat32 -> vfat
[18:05] <cjwatson> I think?
[18:05] <cjwatson> oh, maybe not, these are parted types aren't they, not Linux types, sorry
[18:06] <cjwatson> barry: asymmetry between the set_grub_options impls - you removed the realpath call from one but not the other
[18:07] <ev> going to make the trip home
[18:07] <ev> back in 30
[18:08] <cjwatson> barry: does this still allow grub to be installed to disk devices?
[18:08] <cjwatson> barry: my reading of the code is that it doesn't ...
[18:10] <cjwatson> yeesh, kvm -usbdevice really is unusably slow
[18:12] <cjwatson> ogra_: user part: http://paste.ubuntu.com/894005/
[18:13] <cjwatson> ogra_: the log seems to be truncated with a load of zero bytes before it gets to the tasksel part, so hard to investigate that
[18:15] <cjwatson> ogra_: as for tasksel, 'tasksel tasksel/first multiselect server, openssh-server' preseeds the first run of tasksel during installation, but oem-config counts more as running tasksel again later (although I realise that isn't strictly true in this case)
[18:16] <cjwatson> ogra_: you could preseed tasksel/tasks instead and that should work here, but that's a bit problematic because it would mean any future manual run of tasksel would also be "preseeded"
[18:16] <doko> jibel: about bug 950282, I thought that was seen on more than one machine?
[18:16] <ubot2`> Launchpad bug 950282 in oem-priority/precise "Installation failing with pop-up "The installer encountered an unrecoverable error and will now reboot."" [Critical,Confirmed] https://launchpad.net/bugs/950282
[18:17] <cjwatson> the problem with using 'tasksel --new-install' (hence tasksel/first) is that it will be prepared to remove tasks you don't select as well as installing tasks you do
[18:18] <cjwatson> actually, tasksel/tasks can't be preseeded anyway
[18:19] <cjwatson> so this is tricky.  do you have a use case that actually wants task selection?
[18:23] <doko> any suggestion which issues to look at?
[18:24] <slangasek> doko: there are some more bugs on the pad to choose from: http://pad.ubuntu.com/ubuntu-installer-sprint
[18:24] <bdmurray> while http://paste.ubuntu.com/894016/ fixes bug 960278 we end up having less information in the debug log
[18:24] <ubot2`> Launchpad bug 960278 in ubiquity "ubiquity crashed with UnicodeEncodeError in debug(): 'ascii' codec can't encode character u'\xe9' in position 55: ordinal not in range(128)" [Undecided,New] https://launchpad.net/bugs/960278
[18:24] <slangasek> bdmurray: why encode as ascii rather than as utf-8?
[18:25] <slangasek> it should be safe to use utf-8 as the authoritative encoding for our log files, I think
[18:26] <bdmurray> aiui that won't worth with sys.stderr
[18:26] <slangasek> I was actually wondering on that bug if we shouldn't be ensuring that ubiquity always starts in a UTF-8 locale?)
[18:26] <barry> cjwatson: definitely fat16 and fat 32; you're right that i need to remove another realpath() call (doesn't hurt, but isn't necessary); should still be installable on disk device because that gets added (as 'default') separately, or did you mean something else?
[18:26] <cjwatson> there's C.UTF-8 now
[18:26] <ogra_> cjwatson, well, this is mainly for the QA tests, i doubt they will run tasksel any time later
[18:26] <cjwatson> barry: any disk device should be selectable here
[18:26] <cjwatson> even if it isn't the default
[18:27] <barry> cjwatson: let me add a test and fix for that
[18:28] <cjwatson> explicitly encoding as utf-8 ought to work with sys.stderr, anyway, even if writing unicode objects doesn't ...
[18:29] <cjwatson> ogra_: hm, there used to be a thing you could preseed to leave out some steps, but I'm having trouble finding it now
[18:30] <slangasek> bdmurray: yes, explicit encode seems to work: http://paste.ubuntu.com/894037/
[18:30] <cjwatson> maybe it got lost in the grand refactoring to plugins
[18:30] <ogra_> well, i really dont think the tasksel bit is an issue
[18:30] <cjwatson> yeah, used to be oem-config/steps
[18:30] <slangasek> though I still wonder if we shouldn't just set the locale so stdout/stderr are implicitly UTF-8
[18:30] <ogra_> at least not for the QA usecase, not sure anyone will ever use preseeding in any other scope with the preinstalled images
[18:31] <cjwatson> ogra_: ok, but we do need them to be able to leave it out for automatability, don't we?
[18:31] <slangasek> ... even though this is contrary to the advice in the video barry linked to :)
[18:31] <barry> cjwatson: do you think i need a gtk ui test?
[18:31] <cjwatson> in the special case of ubiquity I think there'd be worse things than running it in C.UTF-8 if that exist
[18:31] <cjwatson> s
[18:32] <cjwatson> barry: depends how hard it is (I forget how much of that exists already); please do if it's not too painful, but I'm not sure it's worth building up a big framework for at the moment
[18:32] <ogra_> cjwatson, well, they would have to mangle the image anyway to set automatic mode in the cmdline and put the preseed file in place
[18:32] <cjwatson> (although if there aren't partman tests, there certainly should be)
[18:32] <barry> cjwatson: ack
[18:32] <ogra_> so the preseed file can as well already have all tasks selected in advance
[18:32] <barry> true
[18:32] <cjwatson> ogra_: the point is that there's no way to do that in their preseed file
[18:32] <barry> slangasek: you could always close sys.stderr and sys.stdout and reopen them with the correct encoding :)
[18:33] <cjwatson> ogra_: I think we might have to add something like 'oem-config oem-config/skip-plugins string tasks'
[18:33] <stgraber> I know we have some partman tests, I had to change/add tests when I worked on the reuse (upgrade) code
[18:33] <ogra_> cjwatson, you think tasksel/tasks wouldnt work ?
[18:33] <cjwatson> ogra_: I know it won't
[18:33] <ogra_> ah, k, i missed that above
[18:33] <slangasek> barry: sick and wrong :)
[18:33] <cjwatson> when I looked at the code I realised that that's explicitly non-preseedable and for good reason
[18:33] <barry> :)
[18:34] <barry> cjwatson, stgraber this one may not be so difficult.  i just need to call set_grub_options() and then check that the menu item has the expected values
[18:34] <cjwatson> barry: there's a build helper script in d-i that has http://paste.ubuntu.com/894042/ near the start
[18:34] <ogra_> cjwatson, yeah, then i think skipping it completely and leaving it to late-command would be best
[18:34] <cjwatson> which I suspect isn't very 3-friendly, but
[18:34] <slangasek> barry: if we're going to do it at all, better to just set the locale before calling ubiquity... but what do you think about that idea in general?
[18:35] <GrueMaster> ogra_: cjwatson:  Would it make sense to have oem-config-remove remove any debconf entries created by the preseed during postinstall?
[18:35] <barry> cjwatson: i was thinking about opening the fd directly to (perhaps) eliminate a layer of calls, but that pastebin will work too
[18:35] <ogra_> GrueMaster, no, you want the dbeconf db to be consistent with your installation
[18:35] <cjwatson> GrueMaster: too hard to figure out I think, and anyway I don't really think it's the right answer here - tasksel has too many problems in this specific context, I really think you just want to skip it
[18:35] <barry> slangasek: that's probably a good thing to in either case :)
[18:36] <slangasek> ok
[18:36] <barry> slangasek: or we could port the code to python 3 now
[18:36]  * ogra_ tries the user-setup bit now
[18:36] <slangasek> barry: sure, then we could just remove the installer from the CD :P
[18:36] <GrueMaster> Well, without tasksel, how are you going to preseed installing ubuntu-server and openssh-server?  Both are needed for testing preinstalled images.
[18:37] <barry> slangasek: our job is done!
[18:37] <cjwatson> GrueMaster: oh, those aren't in the preinstalled image?
[18:37] <slangasek> bdmurray: does using .encode('utf-8') work?  do you think that's a better way to go here?
[18:37] <cjwatson> wah
[18:38] <GrueMaster> cjwatson: They aren't in any image that I am aware of.
[18:38] <ogra_> ubuntu-server surely is
[18:38] <ogra_> but yeah. openssh-server isnt installed
[18:39] <GrueMaster> ogra_: I thought our preinstalled server mirrored the ubuntu-server seed?
[18:39] <ogra_> ubuntu-server must be pulled in by the seed at image build time
[18:39] <cjwatson> ok, let me rethink this a bit then
[18:39] <ogra_> if not, thats a bug
[18:39] <bdmurray> slangasek: I believe it would but should test to be sure.  However this is only one usage of sys.stderr that would be fixed right?
[18:39] <ogra_> likely in the image build system somewhere
[18:39] <cjwatson> in any event I think you'd have to duplicate any tasks in the preinstalled image in your preseeded tasksel/first to make sure that tasksel --new-install doesn't remove them
[18:40] <GrueMaster> I know when I select it during oem-config, I get some 40+ packages installed.
[18:40] <cjwatson> (it was designed to be called on top of a base system, not something that already had some tasks installed)
[18:40] <slangasek> bdmurray: yeah... if we can set LC_CTYPE=C.UTF-8 in the environment everywhere we launch ubiquity, we should be able to fix all of them in one go
[18:40] <ogra_> well, lets just skip that and use late-command
[18:40] <ogra_> its not like this is hard
[18:40] <cjwatson> but skipping it requires infrastructure in ubiquity too; I'd almost rather provide a way to do this in the tasksel plugin
[18:40] <cjwatson> less intrusive
[18:41] <ogra_> k
[18:41] <cjwatson> plus I just started writing the relevant code :)
[18:41] <ogra_> lol
[18:41] <ogra_> you seriuosly are to fast
[18:41] <ogra_> hmpf, still getting the fullname question
[18:43] <ogra_> grrr and i cant get more logging out of it, if i let the install finish i lose my image and have to start over
[18:45] <cjwatson> ogra_: something like http://paste.ubuntu.com/894064/
[18:46] <cjwatson> can I at least have the debug log you have so far from that?
[18:46] <ogra_> its exactly the same as on the bug
[18:46] <ogra_> truncated at exactly the same place :/
[18:47] <ogra_> my prob is that i cant actually properly shut down the board so i cant unmount, only reset or poweroff physically at some point
[18:47] <ogra_> which seems to truncate
[18:47] <cjwatson> sync first?
[18:48] <ogra_> how ?
[18:48] <ogra_> i have no access
[18:48] <ogra_> (serial install)
[18:48] <cjwatson> oh
[18:48] <ogra_> i'll add the new code snippet and let it run through now
[18:49] <cjwatson> I don't know what else might be wrong with user preseeding; that's the only thing I can see that might be killing it
[18:49] <ogra_> could the resetting of the uid be at fault btw ?
[18:49] <cjwatson> no
[18:49] <ogra_> k
[18:50] <cjwatson> that's there to undo the hack in debian/oem-config-udeb.postinst, and otherwise (for non-d-i installs) should be a no-op for pretty much everyone in practice
[18:51] <slangasek> bdmurray: so to go the LC_CTYPE route, it should be sufficient to add 'env LC_CTYPE=C.UTF-8' to the top of each of the upstart jobs in ubiquity, I think; when run from the desktop launcher I would expect it to already have a reasonable locale setting - but I guess that also needs to be checked
[18:51] <ogra_> right, i was just wondering if user-setup gets unhappy if there is no value and asks because of that, but if its a no-op, then it should be fine
[18:53] <cjwatson> no, it's an internal-use hack
[18:53] <cjwatson> won't be asked
[18:53] <ogra_> k
[18:53] <cjwatson> slangasek: why LC_CTYPE and not LANG?
[18:53] <cjwatson> might as well have a default for all the categories
[18:53] <slangasek> bdmurray: hmm, though in the original bug report, it doesn't look like the 'debug' option was passed on the kernel commandline... maybe this was launched manually from the desktop and didn't get a good env
[18:53] <slangasek> cjwatson: fair enough
[18:53] <cjwatson> slangasek: and maybe just do it in bin/ubiquity instead?
[18:54] <cjwatson> os.environ.setdefault('LANG', 'C.UTF-8') or some such
[18:54] <slangasek> cjwatson: requires closing/reopening stdout/stderr after setting it
[18:54] <slangasek> if you expect it to have an effect for the current process :P
[18:54] <cjwatson> could re-exec :-)
[18:55]  * slangasek whimpers
[18:57]  * ogra_ is happy he has LEDs on the panda ... after netcfg i get no output at all anymore now but it seems to be busy doing stuff
[18:59] <stgraber> ogra_: you just need a /dev/console -> morse code converter and you'd even know exactly what's happening :)
[18:59] <ogra_> heh
[18:59] <stgraber> ogra_: then use one of the button as an input device also accepting morse code ;)
[19:00] <ogra_> haha, i wonder if anyone ever did that as a dev project ... TI pays for these
[19:00] <ogra_> or at least awards the coolest one
[19:01] <ogra_> hmm, but it would be nice to get a login prompt at some point
[19:02] <cjwatson> nCipher (where I used to work) had an emergency-use error LED on its hardware security modules that flashed error codes in Morse
[19:02] <ogra_> haha, cool
[19:02] <cjwatson> usually "SOS <letter>"
[19:02] <stgraber> nice :)
[19:03] <cjwatson> given the requirements on what you're allowed to send in and out of an HSM, emergency use actually means "quite often"
[19:03] <GrueMaster> iirc, both leds on the panda along with the second button are user-programmable (although I could easily see someone starting a morse code message on the reset button by mistake).
[19:04] <ogra_> yup, they are
[19:04]  * ogra_ really starts to mistrust this install though 
[19:04] <cjwatson> is there a hardware speaker?
[19:05] <cjwatson> those are great for higher-bitrate encoded debugging
[19:05] <ogra_> the SD occasionally gets accessed and the NIC blinks from time to time as well but it doesnt seem to finish
[19:05] <ogra_> no, there is only a headphone jack
[19:05] <GrueMaster> No, but I'm sure it wouldn't be too difficult to wire one through a gpio.
[19:05] <cjwatson> I've done that while trying to bisect where GRUB's first stage was failing given an extremely limited amount of spare bytes to play with
[19:06] <cjwatson> in fact I think I spent much of one rally with the hoverbook playing loud tunes at me
[19:06]  * ogra_ grins
[19:06] <ogra_>   File "/usr/lib/ubiquity/plugins/ubi-tasks.py", line 39, in prepare
[19:06] <ogra_>     if 'UBIQUITY_AUTOMATIC' in os.environ:
[19:06] <ogra_> NameError: global name 'os' is not defined
[19:06] <ogra_> oh my
[19:07] <stgraber> :)
[19:07] <cjwatson> sorry, http://paste.ubuntu.com/894098/
[19:07] <cjwatson> my bad :-/
[19:07] <ogra_> my bad for not seeing it °!
[19:07] <cjwatson> should've run pyflakes
[19:08] <ogra_> well, fixed
[19:08] <ogra_> in any case that got me more log
[19:08] <stgraber> speaking of pyflakes
[19:09] <CIA-32> ubiquity: stgraber * r5305 ubiquity/ubiquity/misc.py: Make pyflakes happy (unused variable)
[19:09]  * cjwatson runs out of ideas for ways to attack bug 684292 reliably without hardware he doesn't have to hand right now, and asks komputes for help
[19:09] <ubot2`> Launchpad bug 684292 in ubiquity "Installing to USB drive (sdb) using automatic partitioning writes MBR to internal drive (sda) without warning or giving option to change" [High,Triaged] https://launchpad.net/bugs/684292
[19:09] <ogra_> heh
[19:10] <doko> heh, I was looking at this too, but was scared away
[19:10] <cjwatson> misc.grub_default works, so it's something that isn't calling that, I guess
[19:10] <cjwatson> but I really need a debug log
[19:10]  * ogra_ listens to internet radio ... and there was just that beautiful female voice so i thought i should look up the name of the singer ... seems its a man :)
[19:10] <cjwatson> the logs on the bug are useless, they're from a system with only one disk so I don't think they can even be related
[19:11] <cjwatson> and if I rig up something to fake a removable device without it actually being attached as (even emulated/virtual) USB, then there's a risk that I'll miss exactly the buggy code path
[19:12] <cjwatson> doko: did you try reproducing it, even, if you have a useful relevant system to hand?
[19:12] <cjwatson> if so we could work on it together
[19:13] <ogra_> cjwatson, i attached the longer log with the python error to the bug too, but the user part doesnt look different in any way
[19:13] <doko> cjwatson, no, I'm scared away because I'm still travelling and that could damage my only machine now. and it's a mac, using refit to boot
[19:13] <cjwatson> ah yes, that's not going to be desperately useful here
[19:13] <cjwatson> too different
[19:14] <doko> but I'll look at it next week
[19:14] <ogra_> right, but at least its not truncated :)
[19:14] <cjwatson> ogra_: mm, I'm stumped, something must be resetting the seen flag on that question before oem-config starts
[19:15] <cjwatson> if it were something in oem-config itself then we'd see it in that log
[19:15] <ogra_> eeek ! i got the language screen again
[19:15]  * ogra_ checks the kernel cmdline ... could it be that it moved forward even though tasksel stopped ?
[19:16] <ogra_> hmm, no, looks proper
[19:18]  * ogra_ runs debconf-set-selections chrooted on the SD to make sure preseeding is reset to the original values
[19:19] <ogra_> bah, language screen again
[19:20] <ogra_> now thats funny, i had to hit enter for language and region but now it completely seems to finish
[19:20] <ogra_> didnt ask the user stuff either
[19:21] <ogra_> i wonder if the FS just got corrupted, i'm treating it really badly since begining of the day
[19:24] <ogra_> *pling* ... and its in oem-config remove (whose colors are causing eye pain on serial)
[19:30] <GrueMaster> Yea, I wish I could find where that color change is.  Not seeing anything in the ubiquity tree that mucks with the colors.
[19:30] <slangasek> which way do the colors change?
[19:31] <GrueMaster> slangasek: see lp747229
[19:31] <GrueMaster> lp:747229
[19:31] <GrueMaster> bug 747229 ?
[19:31] <ubot2`> Launchpad bug 747229 in ubiquity "weird color change during oem-config debconf package removal step in serial installs" [Medium,Confirmed] https://launchpad.net/bugs/747229
[19:31] <ogra_> slangasek, it looks like the oem-config terminal simply doesnt properly clean up after itself, running oem-config-remove manually at any point doesnt have that effect
[19:31] <GrueMaster> there we go.
[19:32] <ogra_> i planned to look at that after we got preseeding running
[19:33] <slangasek> ogra_, GrueMaster: heh, ok
[19:33] <GrueMaster> I seem to remember something needed to be changed (maverick timeframe iirc) when the changed the default terminal color map.
[19:33] <slangasek> I have no idea on that one
[19:33] <ogra_> nobody has, and its really hard to debug
[19:34] <ogra_> GrueMaster, no, it was there before it just got more obvious with the color change
[19:34] <ogra_> before it wasnt hurting :)
[19:34] <GrueMaster> ok.
[19:35] <GrueMaster> I wonder if it changed when ubiquity changed from unicode to utf-8?
[19:35] <ogra_> (it was black/white inverted back then)
[19:35] <GrueMaster> (pure speculation).
[19:35]  * ogra_ doesnt know when that was
[19:36] <ogra_> ok, starting over with a fresh card and all hacks in place
[19:36] <ogra_> to be sure the FS isnt corrupt etc
[19:37] <ogra_> i also wonder if we had the re-set user stuff because the debconf lock was still in place due to me pulling out the card without unmounting
[19:37]  * ogra_ gets fresh coffee while the card gets written
[19:39] <slangasek> GrueMaster: not sure what you mean by "changed from unicode to utf-8"... did you mean ascii rather than unicode?
[19:40] <GrueMaster> Possibly.  Could just be unintelligent ramblings.
[19:40] <GrueMaster> I really don't know the code that well.
[19:41] <slangasek> well, the only encoding of unicode that's ever been used was utf-8 :)
[19:41] <GrueMaster> Nothing in the changelogs to suggest that though.
[19:42] <GrueMaster> Not correct.  I found where one python script had unicode throughout, and was changed to utf-8.
[19:43] <GrueMaster> ubiquity/plugins/ubi-console-setup.py was where I saw the changes.
[19:43] <GrueMaster> (not that they are relevant to this particular issue).
[19:44] <slangasek> GrueMaster: sorry, I don't know what you saw, but the only unicode encoding we've *ever* used in Ubuntu is utf-8
[19:45] <slangasek> ah, I guess revno 5114 maybe?
[19:45] <slangasek> console interaction would've always been in utf-8
[19:46] <slangasek> but python libraries may use one of two object formats internally, sure
[19:46] <GrueMaster> 5114?  I only have 492 (I could be in a different tree - lp:ubuntu/ubiquity).
[19:47] <slangasek> yeah, I'm on lp:ubiquity :)
[19:48] <GrueMaster> At any rate, I thought I'd throw my eyes into the mix to see if I could spot something, but nothing is coming up.
[19:48] <cjwatson> GrueMaster: I think what slangasek is trying to say is that you have a category error: UTF-8 is an encoding of Unicode
[19:49] <GrueMaster> Ok.
[19:49] <cjwatson> perhaps not to be confused with the Python built-in function unicode()
[19:49] <stgraber> bug 961543 is a weird one :)
[19:49] <ubot2`> Launchpad bug 961543 in ubiquity "menus leave artifacts on screen" [Undecided,New] https://launchpad.net/bugs/961543
[19:49] <cjwatson> but changing from that to misc.utf8 was purely a matter of adding more safety; if you look at the implementation of misc.utf8, it goes like this:
[19:49] <cjwatson> def utf8(s, errors="strict"):
[19:49] <cjwatson>     """Decode a string as UTF-8 if it isn't already Unicode."""
[19:49] <cjwatson>     if isinstance(s, unicode):
[19:49] <cjwatson>         return s
[19:50] <cjwatson>     else:
[19:50] <cjwatson>         return unicode(s, "utf-8", errors)
[19:50] <cjwatson> so it's actually a wrapper around unicode()
[19:50] <cjwatson> lp:ubuntu/ubiquity is the auto-imported tree, none of the developers use it
[19:51]  * GrueMaster never claimed to be a developer.  :P
[19:54] <cjwatson> heh
[19:54] <cjwatson> still, easier for archaeology too
[19:56] <GrueMaster> True.  I'm just wondering if the color issue is even in oem-config.  oem-config-remove looks fine on it's own, and I have not seen anything that would be tweaking the colors during startup.
[19:56] <cjwatson> sounds like the sort of thing you might get from something being run in bterm vs. outside bterm, maybe
[19:58] <ogra_> yeah, i read some of the bterm docs, seems you cant run a bterm inside a bterm
[19:58] <ogra_> i wonder if thats what is happening here
[19:59]  * ogra_ curses
[19:59] <GrueMaster> Except none of /bin/oem-config* runs in a bterm.  Only bin/ubiquity.
[19:59] <ogra_> so this install went straight through to tasksel
[19:59] <ogra_> i wonder if i missed something
[20:00] <GrueMaster> Only the tasksel stuff (hopefully).
[20:01] <ogra_> hmm, no
[20:01] <ogra_> all in place
[20:02] <bdmurray> Does http://paste.ubuntu.com/894178/ look good for a merge proposal?
[20:09] <slangasek> bdmurray: looks good to me
[20:09] <slangasek> cjwatson, ev: ^^ is this worth adding a test case for?
[20:10] <slangasek> I guess that could be relevant for python3 porting
[20:10] <cjwatson> I had half a python3 branch lying around somewhere but it wasn't urgent
[20:10] <ev> yes please
[20:11] <cjwatson> it looks OK but do please test it with both str and unicode types of message
[20:11] <cjwatson> it might well get either
[20:11] <cjwatson> it can just be a unit test, you don't need to construct a full-blown FilteredCommand to do it
[20:12] <cjwatson> maybe with a mock sys.stderr object that checks what gets fed to .write()
[20:14] <slangasek> should that be a new file under tests/, or is this a logical addition to one of the existing ones?
[20:17] <cjwatson> new file
[20:17] <cjwatson> test_filteredcommand.py
[20:18] <cjwatson> mock.patch('sys.stderr')
[20:18] <cjwatson> you can use that as a method decorator
[20:18] <cjwatson> so something like:
[20:18] <cjwatson>     @mock.patch('sys.stderr')
[20:18] <cjwatson>     def test_debug_unicode(self, mock_stderr):
[20:18] <cjwatson>         ...
[20:18] <ogra_> hmm, so the tasksel step only works if i actually use tasksel/tasks
[20:19] <ogra_> (well, not sure yet it "works", but at least it skips it then)
[20:24] <ogra_> ok, dropping the tasksel line from the preseed file makes everything work fully automatic
[20:24] <cjwatson> odd, but ok ...
[20:25] <ogra_> yeah, installing openssh-server can go into late command or so
[20:25] <ogra_> and everything else should be ssh'able then, even running "tasksel install foo"
[20:26] <ogra_> cjwatson, did you do the code changes inb the branch already ? else i have them all here and can commit
[20:26] <cjwatson> no, please go ahead and commit whatever you've tested
[20:26] <ogra_> ok
[20:27]  * infinity wanders through to cough on everyone on his way to find soup.
[20:27] <cjwatson> (FWIW, I'm happy for anyone on the foundations team to have direct commit access to ubiquity, which is just a matter of adding you to the ubuntu-installer team; the reason I don't just add the whole team is that ubuntu-installer gets a LOT of bug mail and people kind of need to have mail filters armed and ready)
[20:28] <infinity> cjwatson: If by "filters", you mean shunting it all to /dev/null, I've been doing that for years. ;)
[20:28] <cjwatson> that's a filter, albeit a trivial one :)
[20:29]  * infinity prefers bugs as an on-demand pull medium, not a push medium.
[20:29] <cjwatson> or maybe "degenerate" is better mathematical slang for this
[20:29] <infinity> Which, unfortunately, does mean I miss out on OMG CRITICAL, DO IT NAO bugs.  But those usually cause a fuss in IRC and the like too.
[20:30] <cjwatson> I filter bugs actually assigned to me differently from ones I'm just subscribed to.
[20:30] <infinity> I could do that.  Still wouldn't solve my criticality issue.
[20:30] <cjwatson> Which seems a decent compromise.  I also arrange for bugs with oem-priority tasks and the like to end up in my inbox.
[20:30] <infinity> Unless I assume someone ELSE is subscribed and reading, and they'll assign to me.
[20:31] <cjwatson> people outside my management chain don't get to say "DO IT NAO", and my management chain knows how to assign bugs :-)
[20:32] <infinity> Actually, come to think of it, assigned bugs DO land in my inbox, while the rest go to .launchdpad-bugs (which may as well be /dev/null)
[20:32] <infinity> But that wasn't intentional.
[20:32] <infinity> I just never looked at why those headers differed enough for one set to slip through. :P
[20:32] <infinity> And it's been like that ever since we switched to Malone.
[20:32] <infinity> I'm *that* lazy. :P
[20:32]  * slangasek assigns infinity ALL the bugz
[20:32] <infinity> GAH.
[20:35] <infinity> cjwatson: I was thinking less about management chain, and more about just noticing quickly.  Like my broken grub2 upload (which you fixed post-haste).
[20:35] <infinity> cjwatson: Granted, that was less about me not having bug mail, and more about me not being near a computer, I think, cause the dude pinged me on IRC and direct email. :P
[20:36] <cjwatson> slangasek: http://ewx.livejournal.com/457086.html
[20:39] <CIA-32> ubiquity: ogra * r5306 trunk/ (4 files in 3 dirs): fix oem-config-firstboot, ubi-tasks.py and bin/ubiquity to make automatic (and preseeding) mode work in oem-config (LP: #924018)
[20:45] <bdmurray> is it just me or my editor? does it look like line 42 in test_timezone.py is missing an ' ?
[20:45] <cjwatson> line 42 is:
[20:45] <cjwatson>         self.gtk.changed(self.gtk.city_entry)
[20:45] <slangasek>         self.gtk.changed(self.gtk.city_entry)
[20:45] <slangasek> ?
[20:45] <bdmurray> 52 then
[20:45] <slangasek> your editor or terminal
[20:45] <cjwatson> probably your editor - it's ' <Unicode heart sign> '
[20:45] <slangasek> u'♥'
[20:46] <cjwatson> without the spaces
[20:46] <bdmurray> right for me the ' overlaps the ♥
[20:46] <ogra_> GrueMaster, see my updates to bug 924018
[20:46] <ubot2`> Launchpad bug 924018 in ubiquity "Preseeding doesn't work with oem-config" [Medium,In progress] https://launchpad.net/bugs/924018
[20:47] <GrueMaster> I see a commit.
[20:47] <ogra_> that too :)
[20:47] <ogra_> i'll test with an apt-get install openssh-server in late command now just to make sure it actually works as expected
[20:48] <ogra_> and with that working we should be done
[20:48] <GrueMaster> I don't think the late command is even triggered.  The sudo changes didn't take affect here.
[20:49] <GrueMaster> Although, I do have the chroot line from netboot, which should be removed.
[20:50] <ogra_> heh, definitely
[20:50] <ogra_> i'll try with a proper line
[20:51] <GrueMaster> Are you going to puch a new jasper with the boot.scr changes?
[20:51] <ogra_> no
[20:51] <ogra_> its up tp the testers to change boot.scr
[20:52] <ogra_> you do it before jasper, it gets carried through
[20:52] <GrueMaster> Oh.  I thought you would do it in jasper as part of the first boot loop.
[20:52] <ogra_> i cant make all server images default to automatic mode ;)
[20:53] <ogra_> jasper just keeps all cmdline entires it doesnt know unmodified and carries them over
[20:53] <GrueMaster> No, but you do already detect the preseed file in jasper during first (resize) boot.
[20:53] <ogra_> that doesnt mean the user wants automatic
[20:53] <GrueMaster> Seems it would be easy to add there.
[20:53] <GrueMaster> Ok.
[20:54] <ogra_> you could set kbd, language and region preseeded but still want the user to use their own PW and name
[20:54] <ogra_> defaulting to automatic every time a preseed file is there would make that impossible
[20:55] <GrueMaster> Yea, didn't think that through.  Multi-tasking atm.
[21:00] <ogra_> i wonder why jasper is so noisy in serial installs ... theoretically all output is redirected to the log
[21:11] <bdmurray> is using EvironmentVarGuard the right thing to do in tests?
[21:11] <cjwatson> if you're setting an environment variable for the use of a single test, yes
[21:11] <ogra_> hmm, no, late-command is actually not executed at all
[21:11] <ogra_> weird
[21:12] <bdmurray> so test_migratiationassistant.py could be chnaged to use it?
[21:12] <cjwatson> bdmurray: it could, though setUp/tearDown is a valid way to do this as well
[21:13] <cjwatson> arguably simpler there since it wants that to apply to all future test cases added to the same class
[21:13] <cjwatson> ev: in testing bug 684292, it's come up that if you have an OS already installed on one disk (say, an internal disk), then you don't get any automatic partitioning options that might allow you to automatically partition another disk (say, a USB stick)
[21:13] <ubot2`> Launchpad bug 684292 in ubiquity "Installing to USB drive (sdb) using automatic partitioning writes MBR to internal drive (sda) without warning or giving option to change" [High,Triaged] https://launchpad.net/bugs/684292
[21:13] <cjwatson> ev: I think there's an argument that this is a weakness in the UX spec; what do you think?
[21:13]  * ev looks
[21:16] <ev> hm, yeah
[21:16] <cjwatson> it'd be nice if you still had the "use entire disk" option
[21:16] <ev> lets point that out to Matthew in the morning
[21:16] <cjwatson> whatever it's called
[21:16] <cjwatson> OK, cool
[21:16] <ev> indeed
[21:17] <barry> cjwatson: okay, i added a ui test.  i think the branch is ready for final review and merging
[21:26] <CIA-32> ubiquity: cjwatson * r5307 trunk/ (4 files in 4 dirs):
[21:26] <CIA-32> ubiquity: Do not offer to install the boot loader on device paths on which
[21:26] <CIA-32> ubiquity: incompatible file systems will be installed (e.g. XFS).
[21:26] <CIA-32> ubiquity: (LP: #959724)
[21:26] <cjwatson> yay, thank you
[21:30] <cjwatson> any more coming for today?
[21:32] <bdmurray> I'm still working on my test and I'm not sure I'll finish today
[21:35] <stgraber> I still "hope" to get some more panel fixes
[21:36] <stgraber> I'm happy to take care of the ubiquity upload once I fixed the bug or gave up :)
[21:37] <cjwatson> ev: could you fix the syntax error I pointed out in bug 930676?
[21:37] <ubot2`> Launchpad bug 930676 in migration-assistant "Ubiquity quits when starting migration-assistent" [High,Confirmed] https://launchpad.net/bugs/930676
[21:38] <cjwatson> it may not fix the whole bug, but it's one thing that's definitely wrong :)
[21:38] <cjwatson> (and I keep running across it in logs)
[21:38] <ev> cjwatson: sure thing
[21:38] <ev> will do now
[21:38] <cjwatson> thanks
[21:39] <cjwatson> stgraber: works for me
[21:43] <CIA-32> migration-assistant: evand * r111 migration-assistant/ (debian/changelog ma-ask): Fix a syntax error (LP: #930676). Thanks Colin Watson.
[21:43] <barry> cjwatson: probably no more from me today
[21:45] <bdmurray> Is there a way to run just one test?
[21:45] <cjwatson> You can run just one test *script*: 'tests/run test_filteredcommand'
[21:46] <cjwatson> the test runner unfortunately isn't very good right now, that's about as granular as it gets
[21:46] <bdmurray> that' good enough for me ;-)
[21:46] <barry> i'd love to integrate a *real* test runner so we can do that :)
[21:47] <cjwatson> can unittest's runner do that, invoked differently?
[21:48] <barry> i'm not actually sure.  i'm mostly familiar with zope.test* which has very nice control over what tests to run.  less familiar with nose, unittest2, etc. etc. etc.  would be nice to investigate a bit
[21:49] <CIA-32> ubiquity: cjwatson * r5308 trunk/tests/run: Allow using a regex as the filter for which tests to run. (Not great, but slightly more useful.)
[21:50] <CIA-32> ubiquity: cjwatson * r5309 trunk/tests/run: remove duplicate import
[21:51] <barry> r5308 looks decent enough for now
[21:52] <CIA-32> migration-assistant: evand * r112 migration-assistant/debian/changelog: releasing version 0.6.11
[21:53] <ev> I caught that the other day, that you cannot specify individual tests
[21:53] <ev> the code for running entire test modules was cargo culted from something pitti wrote, probably apport
[21:54] <ev> but I believe, without digging into the code to back this up ;), unittest parses argv itself, and looks for TestClass.test_function_name
[21:55] <ev> of course my brain is mush from chasing a memory corrupting bug all day, so do ignore everything I say
[21:55] <ev> corruption
[21:57] <ev> or maybe that was something testrepository/testtools was doing
[21:57] <bdmurray> slangasek: so my test is failing - http://pastebin.ubuntu.com/894334/
[21:58] <slangasek> bdmurray: what does the test's source look like?
[21:58] <bdmurray> http://paste.ubuntu.com/894336/
[21:59] <cjwatson> should that not be  variant = u"Arménien"  ?
[21:59] <slangasek> yes
[22:00] <slangasek> bdmurray: also, I think you want filteredcommand.UntrustedBase.debug("Unknown keyboard variant %s", variant)
[22:00] <slangasek> to let the debug function itself do the string composition
[22:00] <cjwatson> agreed
[22:01] <bdmurray> ah, that makes sense
[22:01] <barry> i'm going to take off in a bit, but this has been so much fun, i'm thinking about working on a few more ubiquity bugs tomorrow too
[22:01] <bdmurray> heh
[22:02] <CIA-32> ubiquity: cjwatson * r5310 trunk/ (debian/changelog ubiquity/misc.py):
[22:02] <CIA-32> ubiquity: Make test_misc pass when run on its own. Apparently
[22:02] <CIA-32> ubiquity: GdkX11.x11_get_default_xdisplay() segfaults if Gtk hasn't been imported
[22:02] <CIA-32> ubiquity: (or possibly something finer-grained); so import Gtk from
[22:02] <CIA-32> ubiquity: misc.set_indicator_keymaps even though we don't actually use it
[22:02] <CIA-32> ubiquity: directly.
[22:04] <bdmurray> okay, so I shelved my patch and the test still passes now
[22:04] <slangasek> hah
[22:04] <bdmurray> that doesn't seem right to me
[22:04] <slangasek> nope
[22:04]  * ogra_ packs his bags ... see you tomorrow 
[22:04]  * slangasek waves to ogra_ 
[22:04] <bdmurray> have a safe trip home
[22:04] <ogra_> lol !
[22:04] <stgraber> ;)
[22:05] <GrueMaster> Ah, the power of the Porche.
[22:05] <bdmurray> http://paste.ubuntu.com/894351/
[22:08] <slangasek> barry: ^^ do you have any ideas on this?  I guess the mock sys.stderr is being excessively utf-8 friendly somehow?
[22:08] <cjwatson> mm, yeah, you might need a mock .write method that encodes to ascii
[22:10] <cjwatson> perhaps mock.patch('sys.stderr.write') instead, make the parameter be mock_write, and set mock_write.side_effect to a function that does the appropriate encoding of its args
[22:10] <barry> slangasek: looking
[22:11] <barry> cjwatson: that sounds about right
[22:12] <cjwatson> basically make it the most hostile likely implementation :)
[22:12] <barry> :)
[22:12] <barry> okay, i'm out.  see you guys tomorrow.
[22:12] <cjwatson> ... lambda msg: os.abort()
[22:12] <cjwatson> maybe not
[22:12] <cjwatson> see you!
[22:12]  * slangasek waves to barry
[22:12] <cjwatson> I'd better head off as well, I think.  The kids are starting some kind of riot.
[22:16] <slangasek> bdmurray: untested, since I'm getting build-deps reinstalled for running the test suite, but maybe this?: http://paste.ubuntu.com/894367/
[22:17] <cjwatson> you missed one instance - perhaps move that side_effect function out an indentation level or two so you can easily use it in both places
[22:17] <slangasek> yeah, sorry - was still prototyping :)
[22:17] <cjwatson> or else move the patching to setUp
[22:17] <cjwatson> (examples of that elsewhere in ubiquity's test suite)
[22:18] <cjwatson> e.g. test_misc.py:GrubDefaultTests.setUp which I wrote today
[22:19] <cjwatson> anyway, gone ...
[22:20]  * slangasek waves
[22:32] <slangasek> AttributeError: 'file' object attribute 'write' is read-only
[22:32] <slangasek> that seems unhelpful
[22:32] <bdmurray> that's what I ran into also
[22:39] <slangasek> I get the same thing when I try to do it cjwatson's way
[22:39] <bdmurray> what's the point of using sys.stderr in the test now if we are changing how write works?
[22:39] <slangasek> well, because sys.stderr is what the code we're testing uses
[22:39] <slangasek> so the test has to somehow emulate sys.stderr usefully
[22:42] <slangasek> I think I'm going to punt on this one for today... let's wait until someone who knows mock better can advise
[22:44] <stgraber> ok, I have a fix/workaround/hack for the indicator corruption bug
[22:45] <stgraber> basically the problem is when dealing with labels, with the easiest way to "fix" it being to do just like lxpanel and not deal with the labels, just with icons
[22:45]  * bdmurray heads off
[22:46]  * slangasek waves
[22:57] <CIA-32> ubiquity: stgraber * r5311 ubiquity/ (debian/changelog src/panel/panel.c): panel: Stop dealing with labels, only use the icons
[23:26] <doko> I'm offline now, maybe back later in the evening
[23:34] <cjwatson> slangasek,bdmurray: well, in that case, carry on mock.patching sys.stderr itself, but then add a write attribute to it
[23:37] <CIA-32> ubiquity: stgraber * r5312 ubiquity/ (debian/changelog src/panel/panel.c): Add a indicator_order list to the panel to lock the position of the known ones.
[23:37] <stgraber> finally!
[23:37] <stgraber> getting ready for release + upload now
[23:42] <CIA-32> ubiquity: stgraber * r5313 ubiquity/ (d-i/manifest debian/changelog): Automatic update of included source packages: migration-assistant 0.6.11.
[23:49] <CIA-32> ubiquity: stgraber * r5314 ubiquity/debian/changelog: releasing version 2.9.32
[23:54] <stgraber> uploaded