=== asac_ is now known as asac [00:21] launchpad down again? this is frustrating. [00:21] Seems to WFM [00:21] down for me since last 2 days [00:21] loading forever, that is. [00:21] works here [00:22] Seems snappier than usual to me. Seems like something weird on your end [00:22] IP addr? [00:22] srid: I suggest #launchpad - they actually have some influence over Launchpad's responsiveness [00:22] we, in general, do not [00:23] damn it, it's the damn opera. launchpad loads fine on safari. [00:23] cjwatson: ok [00:26] Crap - typed too quickly. Can somebody unsub ubuntu-motu from bug #394398? [00:26] Launchpad bug 394398 in open-iscsi "Logic to determine expected number of running session wrong (regression in hardy's open-iscsi 2.0.865-1ubuntu3.1)" [Undecided,New] https://launchpad.net/bugs/394398 [00:26] one sec [00:26] ok, it's gone [00:26] Thanks [00:26] LP still isn't snappy here :) [00:27] it has to travel through quite some tubes to get to you though [00:28] ITYM they have to make room on the big truck :) [00:31] hey all [00:33] anyone here that can give me some help with customizing the installer via a preseed file? [00:38] ebroder: it's more like each packet gets hand-wrapped & taken to NZ on a rowboat [00:40] merkur2k: I can, but I'm just about to go to bed. I suggest asking on #ubuntu-installer during European working hours [00:40] merkur2k: and, if you haven't already, read the preseeding appendix in the installation guide (linked from help.ubuntu.com) first [00:41] ok, it might be a simple one. I am wanting to run a fairly complicated script from preseed/late_command that needs user input, but the installer has its own dialog up that blocks the screen [00:41] you'll need to use debconf if you want to interact with the user; not simple ;-) [00:41] just looking for suggestions or ideas for fixes or alternate methods [00:42] it is too interactive to be able to ask questions in advance [00:42] can talk you through the issues and possible fixes tomorrow or whatever [00:42] ok [00:42] debconf doesn't necessarily mean in advance [00:43] at the very least it would just be nice to have some progress feedback other than "running preseed command" [00:43] :) [00:44] but have a good sleep, i will be around [02:39] Given an arbitrary source package (e.g. darcs) and release (e.g. feisty), how can I find out if there is a PPA or backport for a recent version of that package for that release? [02:40] Google [02:40] That is plan B. [02:40] I had the impression that Launchpad had some sort of predictable naming scheme for PPAs [02:40] Not really. They're arbitrary [02:41] Bummer, thanks. [02:41] People creating PPAs for backports usually create a team with a name similar to the package, e.g. https://launchpad.net/~git-core [02:41] But that's entirely at the maintainer's discretion [06:53] Good morning [06:54] Morning pitti [06:55] pitti: So, lool wanted me to fix clutter to build a .pot. I've added langpack.mk to it's rules file, is that all I have to do? [06:56] StevenK: that should do; build log will tell :) [06:58] pitti: But the .pot doesn't turn up anywhere? [06:58] pitti: I can pastebin the build log, if you wish [06:58] StevenK: then it apparently doesn't use standard intltool? [06:59] StevenK: I just wonder why clutter has translatable strings in the first place; isn't it just a graphics lib? [07:00] pochu: the total ddeb archive size is 164 GB, 6 arches, 5 releases [07:00] * pitti cleans up gutsy [07:01] * StevenK waits for fbreader to finish building [07:01] pitti: I'm not sure about a delta from Debian for the removal of -qt and -maemo [07:01] * ScottK decides not to wait for kdesdk to build and goes to bed. [07:02] StevenK: lool reviewed clutter and didn't complain about a missing pot [07:02] pitti: See later comment [07:02] ah [07:05] laga: devkit-power is already used, yes; pm-utils needs to grow support for quirks, it's on the TODO list (https://wiki.ubuntu.com/Halsectomy) [07:05] laga: most devices today shouldn't need quirks any more, but it's still needed for old hw [07:06] good morning [07:12] pitti: http://paste.ubuntu.com/216679/ [07:15] StevenK: looks good [07:15] hey dholbach! *hug* [07:16] * dholbach hugs pitti back [07:16] hi thekorn [07:17] hey dholbach [07:17] hi pitti. so with the old HAL world, I was able to prevent partitions from being exposed in nautilus as user mountable with a nice FDI file. with devicekit-disks, does similar functionality carry over? [07:19] superm1: yes, there is; this is just being re-added, see https://bugs.freedesktop.org/show_bug.cgi?id=22707 [07:19] Freedesktop bug 22707 in detection "Set DKD_PRESENTATION_HIDE for unwanted partitions" [Normal,Assigned] [07:20] superm1: also, bug 394088 [07:20] Launchpad bug 394088 in devicekit-disks "Ignore EFI partition on Intel Macs" [Undecided,Fix released] https://launchpad.net/bugs/394088 [07:20] pitti, ah spectacular. [07:20] yeah i was gonna point out that dell utility partitions are getting mounted and shown too which brought it to my attention [07:20] superm1: our /lib/udev/rules.d/95-devkit-disks.rules already has an earlier version of that patch [07:20] which reading through is already in your comments from a few days ago [07:21] superm1: could you try the patch on the fd.o bug and verify that this covers it? [07:21] you can just apply it to /lib/udev/rules.d/95-devkit-disks.rules and reboot or udevadm trigger [07:21] pitti, yeah i should be able to tomorrow morning [07:22] unfortunately I wiped my recovery partition in a sad accident [07:22] well the utility partition and recovery partitions are different beasts though [07:23] tried to dd an image to my USB stick and accidentally picked sda.. [07:23] after things are sorted for the utility partition, i'm planning on having a third party rule that will be part of the recovery media creation tool on factory shipped ubuntu and rhel6 systems [07:23] ah that'd do it [07:25] superm1: that would work; you should put it into /etc/udev/rules.d/ [07:26] pitti, well it comes part of an installed system in a proper deb or rpm [07:26] ah, ok; /lib then [07:28] normally i'd say it's best to just put it in that upstream devdisk-disks.rules, but unfortunately the partition naming scheme clashes with that of the one for windows factory shipped systems and would cause people with a windows factory shipped system to not have access to their data from linux in a dual boot scenario [07:33] pitti: So, I've done everything for fbreader aside from the i18n, since I have no idea how to sprinkle in a standard intltool build system [07:34] StevenK: that'd indeed be a more intrusive patch, and you should really check with upstream first that they'd take it [07:34] pitti: Right. [07:35] pitti: So does fbreader have a hope of getting approved without i18n? [07:35] StevenK: well, if you need it in mobile, sure [07:35] but as I said, it's really an ugly wart [07:35] and I really recommend fixing it [07:36] sure, it's a day of work, but if you consider it an important use case, it's well worth it [07:36] the intltool build system is easy to do [07:36] most of the work is writing some python code to generate .po files from the existing xml files [07:39] superm1: just uploaded a new dk-disks with current ignore rules [07:40] (needed for NBSing out libsgutils2 anyway) [08:09] pitti: fbreader uploaded [08:12] thanks [08:12] pitti: Bug task slammed back to New, too [08:58] StevenK: +include /usr/share/cdbs/1/rules/langpack.mk [08:58] StevenK: that's not mergeable in Debian AFAIK [08:59] StevenK: You could have used -include foo to the same effect, but that would have been mergeable in Debian; I fixed clutter/clutter-gtk in Debian's pkg-gnome SVN to include gnome.mk instead of autotools.mk which should pull in langpack.mk automatically [08:59] ( ^ confirmed) === tkamppeter_ is now known as tkamppeter [09:23] lool: Okay, I'll fix them both soonish [09:34] is my system supposed to drop out of usplash after finding the initramfs with the last kernel update or did i just see a race ? === luisbg_ is now known as luisbg [10:26] hi asac === ara_ is now known as ara [10:27] hi [10:34] hello [10:35] do ppa's and archives use different buildd's? [10:35] coolbhavi: yes [10:38] Nafallo, please take a look: package is building on my ppa https://edge.launchpad.net/~bhavi/+archive/bhavi-upstream-testing/+build/1116212 but ftbfs on archive buildd: http://launchpadlibrarian.net/28954351/buildlog_ubuntu-karmic-i386.libselinux_2.0.82-1ubuntu1_FAILEDTOBUILD.txt.gz [10:38] coolbhavi: ehrm. why would I be a good person to look at that? [10:38] pitti: quick reply to your mail [10:38] pitti: the Mini 9 has an SSD, I/O seek and load times are _rarely_ an issue [10:39] Nafallo, okay anyone please guide am confused now.. [10:39] Keybuk: well, cold vs. hot cache makes a _tremendous_ difference for gnome-panel, so I wondered how big of an impact it has on ssd [10:40] pitti, most of the gnome-panel start time on my laptop is the menu building [10:41] ie it starts in 1.5s rather than 6s when there is no .desktop installed [10:41] seb128: right, I aked Key to time gnome-menus-ls.py with cold and hot cache [10:41] for me that's 10s (cold) vs. 0.1 (hot) [10:41] seb128: that's what I discussed with vuntz on LinuxTag [10:41] pitti: will do that this morning :-) [10:42] Keybuk: it's not urgent (I won't actually get to this in the next weeks), but I'm curious [10:42] Keybuk: thanks [10:43] pitti, can you please help me? /me scratches his head [10:43] coolbhavi: I guess your PPA package was built earlier, and a new kernel broke it or so? [10:44] pitti, okay now what to do? [10:45] coolbhavi: I didn't look into it yet, it probably nees a small fix for current linux-libc-dev [10:45] pitti: I'm doing the i386 vs. i586 performance tests this morning, easy enough to slip that one in as well [10:45] Keybuk: oh, does that make a significant difference? [10:45] pitti: I'll tell you when I've done the tests [10:45] good luck [10:45] it certainly did with the last set of test CDs we made [10:45] but we think we made those wrong :p [10:46] the last test proved that the new gcc + i586 makes a much faster install than the old gcc + i386 ;) [10:46] pitti, okay, in the archive or in the source that I kept for sponsoring? [10:47] coolbhavi: in libselinux [10:47] hey coolbhavi [10:47] coolbhavi, thanks for your work on the pango update [10:48] okay pitti hey seb128 no mention [10:48] pitti, sorry for the mistake I created [10:48] coolbhavi: don't worry; it wasn't a mistake, just a timing issue [10:48] these things are perfectly normal in a dev release which changes as fast as Karmic [10:49] coolbhavi: the original version would be FTBFS as well now, so it's a completely new issue [10:50] I changed the cflag thing and tested it on my PPA it built so I created a diff [10:52] pitti, now also I tested out on my PPA its building fine [10:52] but I didnt think it would ftbfs in the archive [10:56] pitti: ok, thanks for the input re devicekit and quirks. i read an email from the pm-utils guy from last year saying that he didn't want the quirks in pm-utils, i guess that has changed then [10:57] laga: well, it has to [10:57] nothign except pm-utils is using them [10:57] pitti, any chance we get the drum sound back into gdm at some point ? [10:57] ogra: I consider that a feature :-P [10:57] * ogra doesnt [10:57] :) [10:58] anyway, would need to be hacked into the simple greeter [10:58] i like that i can hear when my system is done booting without watching the screen [10:58] but DX has some plans for this greeter anyway, I think [10:58] and i gues TheMuso likes that too :) [10:58] *gues [10:58] s [10:58] pfft [11:00] ogra: Yeah I have been meaning to look into that. It may be a matter of tweaking the default sound theme for the gdm session. [11:00] yeah, i bet its not to hard [11:00] ah, indeed [11:01] I keep forgetting that it's a normal gnome session now [11:01] pitti: I made the same remark to kirkland about the cron job which should move to PAM (slangasek was complaining about the usage of a cron which spams syslog° [11:01] pitti: I think the issue was that you need to update more frequently than on each login, perhaps to display up-to-date information when displaying it within screen for instance [11:02] After all, pulseaudio gets run by GDM, and if you press the backspace key in the password field, you get a sound. [11:02] pitti: But I didn't confirm that; perhaps this information should simply be updated whenever it's needed rather than regularly though; I agree [11:02] lool: but then screen should directly display the load [11:02] instead of jumping through the python/cron -> motd -> read motd hoop [11:03] a static file is simply not the right thing for this kind of information which updates every second [11:03] (also, why do we need a load, etc. in motd in the first place?) [11:04] pitti: It's not just the load, it's a bunch of things like e.g. number of packages which are update-able [11:04] lool: right, but that can be updated in the apt cronjob [11:05] it just feels like throwing a lot of resources into trying to feed dynamic information into a static file [11:05] pitti: I don't want to dive into specifics; I made the same high level remarks than you did, but I don't use it [11:06] To answer this fully, I would have to check every plugin and see how it could be made efficient [11:06] I agree with you on APT updates and on load; I don't know what else is involved [11:28] pitti: replied with numbers [11:28] Keybuk: wow, SSD has quite an effect [11:29] thanks [11:29] SSD has the effect of eliminating "hot cache" issues from your performance tests [11:29] it's a worthwhile optimization for spinning disks still [11:29] since it's almost as fast to read from the SSD as it is from the page cache [11:29] sure [11:29] but that's not the problem I see [11:29] the problem I see is that there are still fundmental performance issues with gnome-panel *after* you eliminate the I/O problem [11:30] it's named libbonobo [11:30] and sync calls [11:31] sync calls wouldn't show up as CPU time [11:32] how much cpu usage do you get? [11:33] I don't have a recent chart to hand, but it's a fair amount [11:34] ok, I've no been looking at that because it was so small on my chart compared to disk use [11:34] but I'm using a regular disk [11:34] and 80% of the gnome-panel start here is ios [11:34] right, when you're using a regular disk, the I/O wait and load times is what will show up [11:34] hm, an autoreconf in gnome-panel adds "SHAVE_INIT(., enable) [11:34] " to configure [11:34] which fails [11:34] seb128: ah, be careful [11:34] don't fall into *that* trap [11:34] does that ring a bell with anyone? [11:35] pitti, no [11:35] pitti: it fails with a syntax error in configure? [11:35] seb128: if a process is waiting for I/O, it'll show up as I/O wait [11:35] james_w: yes [11:36] even if it's caning the CPU at the same time for something else [11:36] pitti: check Makefile.am has "ACLOCAL_AMFLAGS = -I m4" [11:36] Keybuk, ah, I didn't know that [11:36] so the "80% is I/O" just means that on a rotary disk, you spend 80% of your time in I/O wait [11:36] james_w: find -name Makefile.am | xargs grep ACLOCAL_AMFLAGS -> nothing [11:36] if you take away the I/O, it often doesn't speed it up by removing that [11:36] you just expose what else it was doing while it was waiting for I/O [11:36] often in another thread, etc. [11:36] Keybuk, ok, it will just unmask the cpu use [11:37] I see [11:37] pitti: if there is m4/*.m4 then try adding that line, it fixed it in other packages [11:37] james_w: to all 25 Makefile.am files? *sigh* [11:37] pitti: just ./Makefile.am should work [11:37] * pitti sobs at autotools breakage the 23948234th time [11:37] pitti: then autoreconf [11:37] pitti, in configure.ac is usually enough [11:38] fwiw, autoreconf tells you that you should do this :-) [11:38] it has "AC_CONFIG_MACRO_DIR([m4])" already [11:38] james_w: autoreconf outputs several hundred lines, admittedly I didn't read them all [11:39] if autoreconf outputs more than a dozen lines, you probably have a problem :p [11:39] pitti, "libtoolize: Consider adding `-I m4' to ACLOCAL_AMFLAGS in Makefile.am." is displayed indeed [11:40] admittedly only if you run with -v and work out that that particular message is the one you should pay attention to [11:40] james_w, seb128: I think that helped; thanks! [11:40] np [11:41] that's the only message from autotools [11:41] the flood is for the documentation which you don't really care about [11:42] yay, gnome-panel configured, building now [11:52] who'd like to run a session on merging at UDW? or about anything else ubuntu-develop-y? [11:52] https://wiki.ubuntu.com/UbuntuDeveloperWeek/Prep still has some open slots [11:52] Hmmm. How do I change CPU frequency scaling? [11:52] StevenK, you cant anymore [11:52] its builtin ... [11:53] ogra: so there is only one governor? [11:53] But I don't want this machine to be running at 2GHz [11:53] at least the governor is hardcoded now [11:53] StevenK: The GNOME Panel frequency scaling applet will let you change things temporarily. [11:53] ogra: it might still be possible to change to governor in /sys [11:54] I wasn't after temporarily, either. :-) [11:54] Is ondemand not sufficient to reduce it? [11:54] StevenK, it is set to performance during initramfs and should switch to ondemand after boot [11:54] check if its at ondemand now [11:54] How do I check? [11:54] if not, file a kernel bug [11:55] Since that's one of the things I'm not clear on [11:56] cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor [11:57] should be ondemand [12:25] pitti: Is there a wiki page on how to make multimedia keys work in Karmic? I've got a Dell Mini 10v that's not very happy in that department and I'd like to look into it. === MacSlow is now known as MacSlow|lunch [12:30] asac, just notifying that I'll be here for the next few hours, in case you'd like feedback on https://bugs.launchpad.net/ubuntu/+source/network-manager/+bug/391040 [12:30] Ubuntu bug 391040 in network-manager "When eth0 is unmanaged, system connections for other NICs aren't displayed nor used" [Medium,Triaged] [12:42] pitti: You'll look at fbreader today? [12:43] lool: So, should I undo the change to clutter-gtk, and fix clutter to use gnome.mk? [12:46] StevenK: I guess it's ok as long as the person doing the clutter/clutter-gtk merges know that the langpack include can be dropped [12:47] StevenK: I had in mind to do a couple of Debian uploads and then ask for syncs, but didn't have time yet; I guess we'll get the changes after next upstream release [12:48] lool: Well, I screwed it up, so I should fix it. [12:51] StevenK: It's not very screwed up, just a bit inelegant; if you like to fix it, perhaps you can do a merge from a snapshot of the SVN packages? === MacSlow|lunch is now known as MacSlow [13:20] StevenK: okay; so mobile team (you, lool, etc.) are fine with broken translations? [13:20] pitti: Which one? [13:20] ScottK: incidentally I just updated https://wiki.ubuntu.com/Hotkeys/Troubleshooting for Karmic last week :) [13:22] lool: fbreader [13:22] lool: it doesn't use gettext, but translated xml files [13:23] pitti: Thanks. I'll have a look. [13:25] cjwatson: I managed to get the debian installer to install my product-check udeb (to check hardware and halt the installer) but the script in lib/debian-installer-startup.d doesn't seem to be called: https://pastebin.canonical.com/19636/ [13:26] cjwatson: am I supposed to tell which script the installer should use in debian-installer-startup.d ? === rmcbride_ is now known as rmcbride [13:30] pitti: That sounds bad [13:32] pitti: lool, StevenK: if the decision is to go with xml translations (I'd prefer gettext as well, but I realise this might require some work and coordination with upstream), I'll tell Ubuntu translators that at least they can translate the xml files and also send them to upstream - or maybe someone from the community might want to have a go at adding gettext support to fbreader upstream [13:32] tseliot: name shouldn't matter. To debug it I suggest booting with BOOT_DEBUG=3, exiting the *first* shell you get (but not the second), and stepping through what /sbin/debian-installer-startup would do by hand [13:32] lool, dpm: adding gettext support and changing the build system doesn't actually sound hard; one just needs to write a script to translate the existing .xml files to .po files [13:34] cjwatson: it's an automated installation and after I choose it, there's no way to exit. Or are you suggesting some CTRL+ALT+F1? [13:35] tseliot: so make it not automated for the purposes of debugging! :-) [13:35] tseliot: i.e. boot the installer with BOOT_DEBUG=3 on the kernel command line to see what it's doing [13:36] tseliot: when you do that, it will give you the shells I mentioned, regardless of any automation [13:36] pitti: I think gettext has some support for XML; but I have no idea whether it applies here [13:37] pitti, lool: do you mean that the most work would be simply in writing a one-time script to convert the existing xml files to PO in order to reuse existing translations? I thought the most (or at least most intrusive) work would be to add gettext support (the code uses string id's instead of the English strings) [13:37] lool: no, that's just intltool [13:37] dpm: right [13:37] dpm: as far as I can see, all strings are defined in a single .c file, in a string_constant = blabla('string') manner [13:38] trivial to wrap that in a _() [13:38] pitti: I think xgettext can read Glade files [13:39] pitti: they are in several files (although I guess that's not important), but I think the problem is that string id's are used instead of the English texts. If we wrap them in _() we're going to get the id's as msgids [13:40] dpm: right, the English strings need to go into that file [13:41] cjwatson: it looks like s37speakup fails with "line 1: cannot open /proc/cmdline: no such file" then debian-installer-startup exits [13:41] hi [13:41] are here ubuntu developers? [13:44] pitti: I'm not sure I can follow. So you're proposing: [13:44] 1) adding gettext support to the code [13:45] 2) replace the string id's in the code by the English strings in the .cpp file where those are defined [13:45] 3) wrap those replaced strings in _() [13:45] right [13:46] Replacing the string ids sounds like a large diff [13:46] that's why I said that this should be cleared with upstream first [13:46] it's much easier to do that one-time (with the xml->po conversion), etc. [13:48] pitti, lool: ok, I understand it now, thanks. For me 1) is the gray area, since I don't know how the xml strings are loaded at the moment and I don't really now how much work would be to get rid of their own special l10n system and replace it by gettext [13:49] dpm: I suppose it's a manday of work in total [13:49] but, it's really only worth it if upstream takes it [13:50] maintaining that large delta forever would be a PITA [13:51] cjwatson: that happens because /proc is still empty [13:55] pitti: lool, StevenK: would anyone from the mobile team be willing to do this work for fbreader if I contact upstream (maybe some Ubuntu translators might want to help as well)? [13:56] StevenK: ^ [13:57] tseliot: sounds as though you ignored my advice to exit the first shell straight away, and use the second one it gives you [13:57] tseliot: the first shell is immediately before /proc is mounted [14:00] cjwatson: I can get to the isolinux (boot) screen and to a second shell in which I can select the automatic installation (which then I customise to pass it BOOT_DEBUG=3). Is it wrong? [14:00] on the isolinux boot screen, press F6 to get to extra options [14:00] add BOOT_DEBUG=3 to the kernel parameters there [14:00] press enter [14:00] wait until you get a Unix shell [14:00] type 'exit' [14:00] you will get another Unix shell [14:00] do your work there [14:05] dpm: I would be, yes. [14:15] StevenK: then I'll contact upstream with the proposal and you on CC. Depending on what they say, I'll then also tell Ubuntu translators, since some of them might be able to contribute to the technical work. Sounds ok? === The_Company is now known as Company [14:19] jdstrand: Just about to upload clamav with your apparmor change. I didn't get to it over the weekend. [14:21] ScottK: oh np. there is no rush on that-- it is just sometime during the cycle [14:22] Right. It was mostly I said I'd do it over the weekend and then didn't. === dholbach_ is now known as dholbach [14:25] cjwatson: F6 doesn't seem to do anything on the isolinux boot screen [14:28] cjwatson: never mind, I got to the 2nd shell [14:28] doko, the debian package has the same upstream version, i thought it was an upstream change [14:29] (re binutils) [14:29] doko, the differentce is that debian builds with gcc-4.3 [14:30] tseliot: ok [14:31] ogra: the debian package doesn't have the patch applied [14:31] oh, i thought it was accepted upstream === evanrmurphy_ is now known as evanrmurphy [14:34] cjwatson: it looks like my script is not in /lib/debian-installer-startup.d/ and this is why it's not called. [14:34] cjwatson: currently I make sure that it's installed by adding the package to the list in pkg-lists/local [14:35] maybe the package is installed after debian-installer-startup [14:38] tseliot: no, pkg-lists is entirely processed at build-time not run-time [14:39] tseliot: I don't know why your udeb wasn't included; check the d-i build log [14:39] cjwatson: ok, thanks [15:08] dpm: Sorry, I was AFK. That sounds great. [15:09] StevenK: np, cool, I'll do that, then. === akgraner_ is now known as akgraner === Keybuk_ is now known as Keybuk [15:52] anyone else seeing timeouts on keyserver.ubuntu.com ? [15:53] on --send-keys, i'm getting "gpg: keyserver timed out" and "gpg: keyserver send failed: keyserver error" [16:33] pitti: yo [16:33] ...ghurt ? [16:34] May the Schwartz be with you [16:34] hey kirkland, good morning [16:35] pitti: howdy howdy [16:35] pitti: so you're hating on update-motd as a daemon :-) [16:35] ion: Mr. Coffee or Mr. Radar? [16:35] kirkland, he's not the only one [16:35] kirkland: well, it's a very large club to slam a small problem :) [16:35] pitti: i can revert the upload, but i need to figure out what an acceptable approach might be to all parties [16:35] liw: lool: right, i appreciate your views too :-) [16:36] * liw rejects the entire notion of automatically editing /etc/motd to start with :) [16:36] kirkland: perhaps we should step back for a while [16:36] kirkland: what is the actual goal? displaying system stats on login? [16:36] kirkland: since I don't think that modifying /etc/motd is your actual goal, it's just one way to reach your actual goal [16:37] kirkland: if you want to display system stats on login, wouldn't that be so much easier to do directly in an optional pam module? [16:37] pitti: i don't think it's about the stats at all, at this point [16:37] pitti: the goal is to provide the system administrator, as well as other packages, a method to dynamically inject interesting information into the MOTD [16:38] modifying a static system configuration file every 5 seconds seems like an approach that can only result in weird hacks? [16:38] kirkland: again, into /etc/motd, or for an user to see when he logs in? [16:38] pitti: as for the system stats, they are inherently always stale (which was part of the motivation for the byobu-approach to delivering such system stats) [16:38] (since the latter can be done with a pam module, and doesn't need motd) [16:39] pitti: fair enough [16:39] pitti: we considered that approach back in August [16:39] kirkland: I'm just trying to understand what you actually want to achieve [16:39] pitti: i'm happy to revisit it though, as it's been almost a year [16:39] and obviously I don't know all the history [16:39] pitti: the objection to pam (actually, it was /etc/profile.d that we examined) was that it could introduce a delay on login, gathering such statistics [16:40] kirkland, what is the goal of modifying motd? [16:40] pitti: particularly the calculation of the updates available [16:40] kirkland: well, that's an one-time 0.1 second penalty, as opposed to the permanent penalty you pay by having a resident python daemon.. [16:40] pitti: i suppose we could handle those particular cases with appropriate caching [16:40] kirkland: well, I'm not saying that you should swing to the other extreme and calculate _everything_ on the fly :) [16:40] kirkland: Concerning available updates, there's a daily cron, perhaps you could hook into that? [16:41] kirkland: e. g. the apt cron job could write that information into a cache, and load etc. is cheap to calculate [16:41] lool: yeah, just catting /var/run/updates-available would be the cheapest alternative [16:41] kirkland: you could still have a motd.d/ with scripts, which are run by that pam module [16:41] and one would just display apt's cache file, the other would output the load, etc. [16:42] but then motd could remain static, cause less headache to admins who modify it (and backup systems), and avoids all the cron/daemon stuff [16:42] cjwatson: it looks like my package is only added to the packages pool (in binary/pool/main/p ) but it's not installed. Maybe I should use anna-install to make sure that the package is installed? [16:43] liw: motd = message of the day, but it's no where near that dynamic in its non-update-motd implementation [16:44] kirkland, I know what motd is, and I know it is semi-useless for providing info to people... so what is your goal in modifying it? :) [16:44] pitti: okay, so drop the cronjobs, drop the daemon; replace it all with a compiled-c pam module that runs every time a user logs in that executes each of the update-motd scripts, concatenating the results [16:45] liw: to make it more useful [16:45] kirkland: for efficiency it could provide some standard stuff itself, such as load [16:45] kirkland: but well, it shouldn't matter that much [16:45] login, with all the profile.d/, ecryptfs, etc. stuff is so slow that it would hardly matter [16:46] kirkland: if that'd be feasible, I would certainly like it much better personally; however, what do _you_ think? [16:46] pitti: you think it's better as a pam module? because it would be like a 5-line script in profile.d ? [16:46] I'm not sure where to report this, but the rafal.ca mirror has a 404 on the update manager kde version 0.111.8 for jaunty [16:46] kirkland: oh, if profile.d/ works, so much the better [16:47] kirkland: right, that already has legal.sh [16:47] kirkland: great, that's even easier :) [16:48] kirkland: so, /etc/profile.d/ubuntu-system-info.sh should be all you need? [16:49] tseliot: putting it in pkg-lists/local should have caused it to be included in your image. anna-install will *not help you at all*, nor will *any run-time methods*. [16:49] pitti: i think it would make more sense to keep sysinfo a consumer of a generic update-motd script in profile.d [16:49] pitti: i've gotten a few random emails from people telling me how they're using update-motd for their unique situations [16:50] kirkland: "sysinfo" is the landscape thingy in your context? [16:50] pitti: but if you're okay with that approach, so am i [16:50] pitti: yes, landscape-sysinfo is one consumer of update-motd [16:50] kirkland: sure, I don't mind one more level of indirection [16:50] pitti: cool, i'll run with this [16:50] pitti: until someone else slams me for that approach :-D [16:51] kirkland: so in summary, it needs an extension to the apt cronjob to cache the package updates, and then just another profile.d/ script? perfect [16:51] kirkland: I'll come to your defense then :) [16:51]