[01:39] <CIA-6> partman-base: cjwatson * r181 ubuntu/ (debian/changelog lib/base.sh):
[01:39] <CIA-6> partman-base: Fix regression from calling sed outside debconf_select's inner loop:
[01:39] <CIA-6> partman-base: remove empty choices, which confused cdebconf. This notably happened
[01:39] <CIA-6> partman-base: when asking partman-auto/choose_recipe with a trailing newline on the
[01:39] <CIA-6> partman-base: choices list (LP: #507059).
[10:40] <ara> hey ev, I have one small question about usb-creator, may I?
[10:40] <ev> ara: absolutely
[10:43] <ara> ev, is there any reason why I can't create a disk with my 4gb usb stick?
[10:44] <ev> No, there shouldn't be.  What happens when you try?
[10:48] <ara> ev, in the list of disk to sue, you know that normally appears /dev/sdb (that cannot be used), and /dev/sdb1 that can. in the 4gb usb stick /dev/sdb1 never appears
[10:48] <ara> ev, btw, the distinction between /dev/sdb (that can never be used) and /dev/sdb1 lacks a bit of usability
[10:49] <ara> ev, and the Format button does not seem to be working either
[10:49] <ara> ev, (I am using the gtk version, btw)
[10:59] <ev> ara: can you pastebin the output of devkit-disks --dump
[10:59] <ev> ara: distinction> indeed, there's an outstanding bug or two for that
[11:00] <ev> ara: format> very odd.  Can you also pastebin your ~/.usbcreator.log?
[11:01] <ara> ev, http://paste.ubuntu.com/356506/
[11:03] <ev> ara: was there nothing in devkit-disks --dump for /dev/sdb1?
[11:03] <ara> ev, no
[11:03] <ev> ah, then as far as usb-creator is concerned, it doesn't exist :).  I suspect this is a bug a big lower down, assuming you have a /dev/sdb1 device node.
[11:03] <ara> ev, looking at the log, I have seen that format was not working because it was mounted (but the gtk application did not give me any error)
[11:04] <ara> ev, so that's a bug (I'll file that one)
[11:04] <ev> ara: much appreciated, thanks!
[11:04] <ara> ev, and when I unmounted it, and clicked format, sdb1 appeared, so I guess now I can use it
[11:05] <ev> lovely
[11:08] <ara> ev, bug 507420
[11:09] <ev> cool, thanks
[11:33] <ev> cjwatson: (bug 506585) looking over r3643, I don't see how that can work.  You preseed the name of the script, but ask_user only accepts responses of the form "80script" or "80script________script", not "script".
[11:36] <cjwatson>                         if [ "$default" = "$plugin" ] ||
[11:36] <cjwatson>                            [ "$default" = "${plugin#[0-9][0-9]}" ]; then
[11:36] <cjwatson>                                 default="$key"
[11:36] <cjwatson>                         fi
[11:36] <cjwatson> ought to arrange for "script" to work ...
[11:36] <cjwatson> (in debconf_select)
[11:36] <cjwatson> I did test that change, I guess not well enough, sorry :-(
[11:41] <ev> ah, sorry about that.  Didn't go far enough down into the rabbit hole, it seems.
[11:41] <ev> completely missed that substitution
[11:42]  * ev digs further
[11:44] <cjwatson> odd
[11:44] <cjwatson> I think I'd suggest the 'set -x' hammer to investigate ...
[11:50] <ev> indeed, thanks
[12:50] <ev> cjwatson: ah!  It doesn't work because that processing (debconf_select) is done just before the question is actually asked, and partman.py only preseeds 'delete' as the question is being asked.
[12:56] <cjwatson> meh
[12:56] <cjwatson> so how do we fix this?  the core problem in the original bug is that with the perf optimisation partman.py doesn't necessarily have the list of scripts in hand
[12:56] <cjwatson> I think perhaps we need to add something to debconf_select to let that kind of preseeding work ...
[12:57] <cjwatson> ... which is tricky 'cos debconf_select doesn't know either, if choices are frozen
[12:57] <cjwatson> maybe this whole frozen choices thing was a bad idea, but it does make things so much faster :-/
[12:58] <cjwatson> I suppose we could have ask_user look for [0-9][0-9]$RET or something ...
[12:58] <cjwatson> in $dir
[13:59] <ev> sorry, had stepped out for lucnh
[13:59] <ev> hrm
[14:38] <superm1> ev, regarding removing oem-config post install, perhaps should you be removing 'ubiquity' instead of 'oem-config'?  removing ubiquity should knock out all of oem-config too
[15:02] <ev> cjwatson: I take it the choices being frozen prevents us from using must_find_one_script on menu_options?  If so, I'll agree for lack of a better solution.
[15:02] <ev> cjwatson: something like this http://pastebin.ubuntu.com/356601/ ?
[15:02] <cjwatson> so the point of freezing choices was to avoid spending so much time listing out the choices when we already knew what we were going to do
[15:02] <ev> err, just written properly, with $RET
[15:03] <cjwatson> the problem is a fencepost error
[15:03] <cjwatson> we thaw the choices one step too late
[15:03] <cjwatson> it might be possible to work out that we're about to return to the top level and return control to the user, and thaw choices then, maybe
[15:04] <cjwatson> that sort of thing was what I was thinking of, although be careful with what test(1) will do if there's more than one match; you might have to use a for loop and break
[15:11] <cjwatson> possibly something like http://paste.ubuntu.com/356608/, not that that makes things any less ridiculously complicated in there
[15:11] <cjwatson> does that make any sense?
[15:12] <cjwatson> either we need to handle bare preseeding as your paste, or we need to arrange to thaw choose_partition choices just before we get to the point where we currently realise that we've finished building the cache
[15:15] <ev> that seems significantly cleaner
[15:17] <cjwatson> as usual the partman component state machine could do with some kind of diagram :-(  I'm not entirely confident that that's the only possible transition back to choose_partition when choices are frozen
[15:17] <cjwatson> although I *think* it is
[15:18] <ev> superm1: good call
[15:19] <ev> indeed, my eyes are slightly sore, having dug through this
[16:06] <CIA-6> ubiquity: superm1 * r3655 ubiquity/ (bin/oem-config-firstboot debian/changelog): Adjust previous commit to remove ubiquity rather than oem-config (ubiquity will knock out of all oem-config too)