CIA-32 | preseed: cjwatson * r941 ubuntu/debian/ (71 files in 2 dirs): merge from Debian 1.50 | 12:27 |
---|---|---|
CIA-32 | preseed: cjwatson * r942 ubuntu/debian/ (changelog network-preseed.templates): Change default preseed root to "d-i/precise/./preseed.cfg". | 12:28 |
CIA-32 | preseed: cjwatson * r943 ubuntu/debian/changelog: releasing version 1.50ubuntu1 | 12:29 |
CIA-32 | iso-scan: cjwatson * r553 ubuntu/debian/ (31 files in 2 dirs): merge from Debian 1.37 | 12:34 |
CIA-32 | iso-scan: cjwatson * r554 ubuntu/debian/changelog: releasing version 1.37ubuntu1 | 12:36 |
davmor2 | ev I think I have found the issue with my laptop and installing Ubuntu, when the slide show starts does it refresh the frame buffer? | 15:09 |
ev | davmor2: it's several layers removed from that. | 15:15 |
davmor2 | ev: are you back in London? | 15:17 |
ev | I am! | 15:17 |
ev | finally | 15:17 |
ev | much as I enjoy the bay area, it was *cold* | 15:17 |
davmor2 | ev: Right I'm down tomorrow with the laptop, so if you want to have a look you are more than welcome once I do my demos | 15:17 |
ev | davmor2: indeed, I've been asked to attend the sprint tomorrow as well | 15:19 |
davmor2 | ev: I have a horrible feeling that it is basically down to the intel hw and the fact that it is an odd fb size ie none standard 1366x768 or something like and I think 1364 is the default off the top of my head | 15:20 |
davmor2 | ev: I want to see your ldtp script in full swing too cause techinically at the moment it should fail ;) | 15:24 |
ev | okay | 15:25 |
stgraber | davmor2: I'll have a look at the d-i code to see exactly what's going on there. For the scanning part, wlan0 seems to be down here so that'd explain it, for the skipping issue, that's likely just a wrong check in debconf | 16:43 |
davmor2 | stgraber: indeed I'm assuming it's something minor | 16:44 |
davmor2 | stgraber: the other nice flaw is if you put in a wireless point and continue with the install you then can't connect to any wireless points because it set a value in /etc/host/interfaces iirc | 16:50 |
davmor2 | stgraber: except obviously for the one you set which works flawlessly with no wifi connection according to nm | 16:51 |
stgraber | yeah, I'd expect netcfg to configure /etc/network/interfaces which makes NM ignore the interface completely | 16:52 |
stgraber | I guess we could try and generate the matching NM configuration file but that may get a bit tricky and I'm not sure we want to do that for 12.04 ;) | 16:52 |
davmor2 | stgraber: indeed something for the future maybe :) | 16:53 |
stgraber | cjwatson: syslog: http://paste.ubuntu.com/870145/ | 17:00 |
stgraber | cjwatson: proposed fix for the scanning loop: http://paste.ubuntu.com/870146/ | 17:00 |
cjwatson | looks plausible ... | 17:02 |
cjwatson | doesn't look like the stuff that asks netcfg/wireless_security_type etc. has any idea what the answer to netcfg/wireless_essid_again was | 17:06 |
stgraber | yeah, I think the issue is that netcfg_wireless_set_essid returns 0 in all cases | 17:09 |
stgraber | so the rest of the code assumes everything went just fine and prompts for the security options | 17:09 |
* stgraber wonders if changing that "break" in from the wireless while loop to a "return GO_BACK" would do the trick | 17:10 | |
stgraber | s/in from/from/ | 17:10 |
cjwatson | maybe, though that doesn't feel like the right error code - it's not "go back", it's "skip forward further" | 17:12 |
stgraber | yeah, GO_BACK seems a bit weird indeed but it's the only return value that netcfg seems to look for when calling netcfg_wireless_set_essid... | 17:13 |
cjwatson | right, but make it look for more things rather than abusing that return value :) | 17:14 |
cjwatson | it's OK to change the surrounding code ... | 17:14 |
stgraber | cjwatson: diff looks something like http://paste.ubuntu.com/870203/ then | 17:38 |
stgraber | (haven't tested yet) | 17:38 |
cjwatson | I'll take your word for it :) | 17:40 |
stgraber | waiting for an i386 chroot to be ready for a test build... for some reason I don't have any amd64 hardware with a wifi chip | 17:44 |
stgraber | ok, apparently moving the up code somewhere else isn't enough, I'm starting to suspect the essid listing code to be completely broken (at least on some hardware) | 18:35 |
stgraber | running "iwlist wlan0" gives me the right results | 18:35 |
stgraber | hmm, looks like I was wrongly assuming it was actually attempting to scan which it's not, it's simply setting an empty essid and hoping it'll select an open one | 18:41 |
stgraber | which might actually work in some cases, so I guess that part wasn't a bug at all, just me not reading properly | 18:41 |
stgraber | so will restrict the fix to just the debconf part which currently kind of works (brings the user back to the main d-i menu instead of the netcfg menu, will have to fix that bit) | 18:42 |
stgraber | cjwatson: http://paste.ubuntu.com/870305/ seems to work here. Basically if the user doesn't enter an essid it's going to exit netcfg (return 0) and continue the installation without networking | 18:54 |
stgraber | my initial idea of going back to the network card selection list wouldn't match the label that says "To skip wireless and continue, leave this field blank." | 18:55 |
stgraber | oh, I probably should use GET_HOSTNAME_ONLY instead of QUIT though, otherwise we end up without an hostname | 18:57 |
stgraber | davmor2: just to confirm, you expect netcfg to just prompt for your hostname and continue with networking in cases where you don't enter an essid, right? | 19:18 |
davmor2 | stgraber: Like it does if you select the ethernet port but don't have a cable connected | 19:19 |
stgraber | well, showing that screen would be a bit weird as you'd essentially end up entering static config for a wireless network without an ssid... I think it makes more sense to jump directly to the "Do not configure the network at this time" which will just prompt for your hostname | 19:23 |
stgraber | the only problem I see with doing that though is if the user wants to re-try the wifi autodetect stuff | 19:23 |
stgraber | but they should be able to do it by choosing "Go back" and selecting the network entry again | 19:23 |
davmor2 | stgraber: no I mean the dialog for hostname sorry | 19:30 |
stgraber | ah ok, right, that's what I have here now so I'll run one more test and upload netcfg with that then | 19:30 |
davmor2 | stgraber: right so if no network is selected in the first one and none found in the scan and you don't set one the second time it moves onto hostname? | 19:31 |
stgraber | yep | 19:33 |
stgraber | davmor2: did you file a bug report? | 19:34 |
davmor2 | stgraber: nope I just noticed it and wondered if it was an issue or me going mad, turns out it was an issue for a change | 19:35 |
stgraber | ;) | 19:36 |
davmor2 | stgraber: hey the amount of bugs I find that no-one else can find and even with screenshot evidence can't reproduce is disturbing | 19:39 |
CIA-32 | netcfg: stgraber * r1294 ubuntu/ (debian/changelog dhcp.c netcfg.c netcfg.h wireless.c): When wireless detection doesn't find any network and the user doesn't enter any essid, skip directly to the hostname dialog rather than prompt for security options on a non-existing network. | 19:41 |
stgraber | davmor2: actually, bug 36426 | 19:42 |
ubot2` | Launchpad bug 36426 in netcfg "Can't skip configuring wireless network" [Medium,Confirmed] https://launchpad.net/bugs/36426 | 19:42 |
davmor2 | stgraber: man that's an old bug | 19:43 |
stgraber | yeah :) | 19:44 |
davmor2 | stgraber: Yay for closing really old bugs that have affected people for 6 years | 19:49 |
davmor2 | I'm assuming not many people are force to use the alternate installer on a laptop nowadays | 19:50 |
cjwatson | what people encounter doesn't necessarily correlate with what we have time to fix | 19:52 |
cjwatson | that would be nice, but ... | 19:52 |
davmor2 | cjwatson: Indeed, I'm assuming it isn't come across so often anymore though with the live installer working so well nowadays, so the d-i would I assume be more server installs in which case they would almost always be etherneted installs | 19:54 |
cjwatson | lvm/raid/crypto still the big ones | 19:55 |
cjwatson | I know quite a few Canonical staff still use the alternate installer on laptops for that | 19:56 |
cjwatson | never mind anyone else :) | 19:56 |
davmor2 | cjwatson: hmmm okay you don't think of it like that I guess, well I don't lets put it that way | 19:56 |
CIA-32 | netcfg: stgraber * r1295 ubuntu/debian/changelog: releasing version 1.68ubuntu13 | 20:41 |
brendand | should the installer be able to select a kernel to install based on the RAM installed on the system? | 21:21 |
brendand | i.e. judge when to use pae? | 21:21 |
cjwatson | yes, but only from the kernels on the medium it's installing from | 21:22 |
cjwatson | a CD-based installer intentionally only installs kernels from the CD | 21:22 |
brendand | hmmm. could pxe installing have anything to do with it? | 21:28 |
cjwatson | a netboot installer should be able to select an appropriate kernel | 21:29 |
brendand | cjwatson, sorry. think i just ran out of battery. | 21:31 |
cjwatson | 21:29 <cjwatson> a netboot installer should be able to select an appropriate kernel | 21:31 |
brendand | ok. we'll investigate more and make sure there's nothing we've missed | 21:32 |
cjwatson | could be a bug of course, I don't think I've tested this specifically since precise switched to the installer running on a PAE kernel | 21:36 |
brendand | cjwatson, this is oneiric. i understand precise is a different story | 21:40 |
CIA-32 | ubiquity: cjwatson * r5231 trunk/ (debian/changelog ubiquity/plugins/ubi-migrationassistant.py): Fix debug message when failing to create the migration-assistant page. | 23:47 |
CIA-32 | ubiquity: cjwatson * r5232 trunk/ (debian/changelog tests/test_gtkui.py): | 23:55 |
CIA-32 | ubiquity: Ensure that the migration-assistant page is shown in | 23:55 |
CIA-32 | ubiquity: test_pages_fit_on_a_netbook. | 23:55 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!