/srv/irclogs.ubuntu.com/2011/01/26/#ubuntustudio-devel.txt

ScottLTheMuso, my attempt at fixing the gnome-classic xsession has failed, i will talk with didrocks tomorrow about it01:39
paultagScottL: roughing out some of the controls work -- this'll be cake01:39
paultagScottL: if I put some time into it over the weekend, I can slam it out with time to have a beer01:39
ScottLpaultag, awesome!  i look forward to seeing it01:40
paultagScottL: righto!01:40
* paultag goes back to work01:40
paultagScottL: ping?03:03
ScottLpaultag, pong03:07
paultagScottL: :)03:07
paultagScottL: Is there a full list of all the things you want to see on this app? I just hunted a graphic designer for this, so we can do a bit more03:08
paultagScottL: https://wiki.ubuntu.com/UbuntuStudio/SettingsApp/Redux03:08
paultagScottL: I've jotted what comes to mind there03:08
ScottLpaultag, this is what we have, it's pretty jumbled but i was cleaning it up:  https://wiki.ubuntu.com/UbuntuStudio/ControlsRedesign03:10
paultagScottL: what about wild-fantisy stuff?03:10
ScottLi was migrating it to there from this one:  https://wiki.ubuntu.com/UbuntuStudio/ReleasePlanning#Ubuntu%20Studio%20Controls03:10
ScottLpaultag, i hadn't really thought that far outside the box to be honest becuase i was worried about getting this much done :P03:11
ScottLbut i thought it might be a good place to do any config stuff for audio03:11
paultagScottL: roger. I'll cherry pick from there. doctormo and I work well together ;)03:11
ScottLmaybe we can brainstorm a few more ideas with holstein, ailo__ , astraljava 03:11
paultag_103:11
paultag+1 *03:11
paultagcheers, thanks ScottL. I'll send you updates as they happen03:14
ScottLpaultag, so what is doctormo doing with this?03:14
paultagailo__: I'm jacking devel on the settings stuff ;)03:14
paultagScottL: He and I will both do GTK. I'll do low(er) level code in python ( settings interfaces, etc ), while he does icons and makes everything cosmetic03:15
paultagScottL: it will let me focus on getting it done, while he bitches about how it's unusable ( until it is ;) )03:16
ScottLpaultag, LOL...but aye, he's good with icons and graphics03:16
paultagScottL: we're great buddies, he and I are both in Boston03:17
ScottLhaving him help is a boon03:17
paultagScottL: we go for tea now and again03:17
paultagScottL: yeah, for sure03:17
ScottLpaultag, are you in a loco?03:17
paultagScottL: Yeah, -us-ohio -- I was contact until two months ago03:17
paultagScottL: I'm also on the loco-council ;)03:17
ScottLi believe he's british, are you british as well, or do you just enjoy tea with a friend :)03:17
paultagScottL: he lives in Boston, MA now ;)03:18
paultagwhich is where I'm from 03:18
paultagso when I go back home, I meet up with doc-mo03:18
doctormohey ScottL03:22
paultagheyya doctormo :)03:22
doctormopaultag is trying to explain to me what options are needed for this new setup tool03:22
paultagScottL: we were just talking in PM, figured we'd drag you into this03:22
doctormohttps://wiki.ubuntu.com/UbuntuStudio/SettingsApp/Redux#preview03:23
doctormoBut the settings as documented appear to be install time settings which should be done there. Not in an extra tool.03:23
paultagdoctormo: iirc one of the issues was a DVD install vs migrating an Ubuntu install via metapackages03:24
paultagdoctormo: you should be able to install ubuntustudo-desktop and get it, and not just have to use the DVD installer03:25
paultagIIRC, that is03:25
doctormoYes, I understand that.03:25
paultagk03:25
ScottLhi doctormo 03:26
* ScottL was talking with the wife03:26
ScottLdoctormo, many people also start with a vanilla ubuntu install and just install what they want/need from ubuntu studio as paultag pointed out03:28
ScottLoh, the package name is actually ubuntustudio-controls03:29
ScottLpaultag, doctormo : the extra software bit probably may not be a good thing at this time, many felt that this was a bit dangerous03:30
paultagScottL: righto. Let's remove that bit,t hen03:31
paultagthen *03:31
doctormoScottL: Are there controls which can/should be toggled? or is this a one way thing?03:31
paultagdoctormo: the rtprio / memlock need to be massaged03:31
paultagdoctormo: usually. but once it's set right, it's OK03:31
doctormoSo that's a set once, never touch again type thing?03:32
paultagdoctormo: if it's set right, yes. If it's set wrong ( and it locks up yoru system ) you need to tweek it03:33
doctormodangerous then eh, I take it you tweak it with a livecd?03:33
paultagdoctormo: I rawdog it. It's tough to cause any real huge issue03:33
paultagdoctormo: just reisub it03:34
doctormolanguage I don't understand!03:34
paultag(if the rtprio is set too high, it sucks all the cycles )03:34
paultagdoctormo: alt + sysreq + {R, E, I, S, U, B}03:34
paultagdoctormo: that resets your machine nicely03:34
paultagdoctormo: but works even with a total lock ( short of a kernel panic )03:35
paultagdoctormo: regardless, I see this as a userspace app, the more I think about it03:37
doctormopaultag: Design wise, it's not. Unless you have further data.03:38
paultagdoctormo: how is it not? What if you add a user?03:38
paultagdoctormo: what if you want to switch from rt to lowlatency later on?03:38
doctormodoes rtprio need to be set for each user?03:38
paultagdoctormo: or change the rtprio03:38
paultagdoctormo: it can be, but it does not need to03:38
doctormoConsider sane defaults with cli tools to control the complexities if required.03:39
paultagdoctormo: again, that won't work for migrating installs03:39
paultagdoctormo: you have to install, set up a user, apt-get install the metapackages, and you can't touch /home in packages03:39
doctormoSure it will, you can have post-inst callouts on a package attached to the meta package to do the work.03:40
paultagdoctormo: no, I mean, /home does not exist for packages03:40
paultagdoctormo: if it touches it, it's far out of spec03:40
doctormoYes you can touch home in packages... technically speaking the current user.03:40
paultagdoctormo: no, you can't03:40
paultagdoctormo: that's wildly against policy03:40
doctormoThen there are a whole host of naughty debian packages.03:40
paultagdoctormo: no there are not03:40
doctormolol, ground control?03:41
paultagdoctormo: in what case does a package touch /home/user/*03:41
paultagdoctormo: that seriously writes stuff to /home/*/* ?03:41
persiaYou can't touch /home in packages, because FHS says /home is OPTIONAL.03:41
paultag+1 persia 03:41
doctormoThe installer restarts nautilus, albeit on the users say so03:41
paultagdoctormo: that's running a command that sends a signal03:41
doctormopersia: I defer to you,03:41
persiaYou can, if you insist, have postinsts that use getent or similar to modify things in user home directories, which is subtly different.03:41
paultagdoctormo: that does not touch /home03:41
paultagpersia: even that should be handled on the application level iirc03:42
persiapaultag, Depends on the package.  For something like ground control, it's better to use the install notify hook to ask the user to restart nautilus (or similar).03:42
persiaUse of wrappers is more common.03:42
paultagpersia: sure, but the app it's self will init the files in ~/03:43
paultagpersia: not the package03:43
doctormoOK chaps, off to bed, paultag figure out what fields you need and update that wiki and note me. I'll have a crack at a mockup for you.03:43
paultagdoctormo: cheers, thanks03:43
paultagdoctormo: tea when I get back ;)03:43
doctormoaye aye03:43
ScottLdoctormo, thank you for the help03:43
persiaif [ -f $HOME/.config/myapp ]; then myapp; else; cp /usr/share/myapp/config $HOME/.config/myapp; myapp; endif03:43
doctormoScottL: Of course, I still owe you some kind of wacom tool, and if it were possible to configure wacom instead of just creating a xinit script with commands in it, it'd be done ;-)03:44
ScottLpaultag, i'm terribly excited to see what you chaps come up with, and i know the users are excited for improvement as well03:44
paultagpersia: yeah, but that looks like a script that's calling myapp -- and can be run when you run the command03:44
persiapaultag, Depends.  Packages that have dedicated users for some reason may need to do it in postinst.  When dealing with real users, yes, better to use a wrapper (better support for later adding users, etc.)03:44
paultagScottL: :)03:44
paultagpersia: aye :)03:44
persiaYeah, that's a wrapper script :)03:44
paultagAhhh, yes03:44
paultagthat's how I do it with fluxbox :)03:44
paultagthanks for the help persia :)03:46
ScottLhi troy_s , i talked to quadrispro about that app you mentioned needed packaging and he's interested but just asked me to remind him later as he's busy at the moment04:26
troy_sScottL: rexbron is almost there I believe.04:27
ScottLtroy_s, oh, that's cool then04:29
quadrisproScottL, we have a big problem with ladish's licensing09:30
quadrispropersia, could you give me an opinion? http://paste.ubuntu.com/558457/09:31
persiaquadrispro, What is the "Academic Free License version 2.1"?11:25
quadrispropersia, http://opensource2.usrbinruby.net/licenses/afl-2.1.php11:26
persiaAnd why do you mention "* This file contains D-Bus methods helpers"?  Are those under separate license?11:27
quadrispropersia, it is solved, upstream meant to cover some portions of the code by dual-licensing11:27
quadrispropersia, libdus is dual-licensed, too11:27
persiaOK.  So, here's my summary recommendations:11:28
* quadrispro hasn't noticed it before, really11:28
quadrisprohttp://packages.debian.org/changelogs/pool/main/d/dbus/current/copyright11:28
persia1) You should identify what files are covered by which licenses11:28
quadrisprodone11:28
persia2) You should make clear that dual-licensed code is dual-licensed11:28
persia3) You should include the *complete* text of afl-2.1, as it is not in /usr/share/common-licenses/11:29
quadrispro2. reported upstream, and he says that it will be clearly stated in the next release11:29
quadrispro3. done11:29
persiaFor 2) if you have that from upstream, and you know which files are licensed how, you can just report that in debian/copyright, rather than waiting for the next release (unless it's very soon)11:30
quadrisproI'd prefer to have a public statement which clarifies that. Upstream seems very collaborative, he's going to publish a small note on the website in the next few hours, too11:34
persiaAh, the small note will be hugely useful, indeed.11:34
quadrisprothough, we need to introduce an important change to jack to make ladish work fine11:35
persiaIndeed, which was a stumbling block for the longest time, and, if I recall correctly, one of the reasons nedko went to work on JACK2 upstream.11:35
quadrisproAFAIK, first of all jack needs to compile with a sort of no-self-connect option enabled11:37
quadrisproto avoid conflicts between apps' self-connecting feature and LADISH's session management11:38
persiaUm.  That doesn't sound ideal: there's been a lot of effort to make JACK "Just Work" by enabling self-connect.  Do you think that we can get enough documentation that users won't be annoyed, or set up some sort of initial auto-connect with ladi?11:41
quadrisprodocumentation is never enough :) and yes, it would introduce several complications11:47
quadrisproI need to send Adrian all information I've collected11:48
persiaSo, what still depends on the "audio" group?  My understanding is that most stuff has evolved to no longer depend on this, and if we can find out what is outstanding, and clear the rest, that is probably better than making sure users have an easy way to be added to the group.  Part of "Just Works" :)13:53
scott-workpersia:  i thought that the real-time time privileges given when adding the jack package depend on being in the audio group14:50
scott-workalso i thought the new udev rules for firewire depend on the audio group14:50
scott-workoi, i thought ubuntu.com and launchpad.com were slow...however, kernel.org must be on dialup in the antartic14:51
scott-workpersia: https://ieee1394.wiki.kernel.org/index.php/Juju_Migration#Permissions_and_ownership_for_.2Fdev.2Ffw.2A shows that it depends on video group but i believe ailo looked into this and found that it was the audiog roup14:52
scott-workof course i could be mistaken14:52
scott-workabogani: good morning!  i notice that the kernel blueprint for natty is complete with one notable exception, getting documentation for kernel derivatives/flavour into the repositories14:53
persiaDepends on the udev rules, really.  Should be possible to consolekitify it, I'd think.14:54
scott-workabogani: hopefully we see something soon and in time for natty14:54
scott-workpersia: i will defer to your knowledge and judgement on that as i know practically nothing about consolekit14:54
persiaMay be best to just upload something, really, and fix it later.  Last time I checked with apw, he said that the team had lost a lot of folk, and was having trouble getting all their work done (which is why the doc isn't published yet)14:54
scott-workpersia: upload kernel to the repository?14:55
persiaSure.14:55
persiaMaybe best to stick with the group for now then.  I suspect it's fixable, but it may be non-trivial, sadly.14:55
scott-workabogani: ^^^ please note that persia is advocating uploading your -lowlatency kernel into the repository, unless i am misreading14:56
persiaIndeed I am.  Waiting further just gets annoying, although without the document, there may be work to do later to bring it into compliance.14:57
scott-workbut waiting longer also risks not getting the kernel into natty14:57
scott-workwhich is bad i should add14:57
persiaIndeed.15:04
ailoThere's now udev rules included for firewire with the standard Ubuntu install in /lib/udev/rules.d/69-ffado.rules which point to Audio group15:12
ailo60-ffado.rules*15:12
persiaIt should really be richer than that.  Audio interfaces should be granted permissions similar to USB/PCI/etc. audio interfaces.15:13
ailoBut those are only for realtime, what I understand15:13
persiaStorage devices should be given extremely restrictive permissions.15:13
persiaVideo devices should be granted permission as needed (I don't have enough knowledge of how video processing works to know if this is restrictive, group video, etc.)15:13
ailoThe firewire device should work without editing, and using no realtime with jackd15:14
ailoWe tested firewire, and this is what we have found, so far15:14
persiaSure, but with juju+udev, we no longer need have a single policy for all devices: we can have a per-device policy, based on what the hardware actually does.15:14
ailoThere is per device policy in the 60-ffado.rules file15:15
scott-workailo: did you specifically test if the user does needs to be in the audio group (i.e. does performance degrade if the user is not in the group?)....i don't current remember if you did or not15:15
ailoBut still, only for realtime, as what I can understand15:15
ailoAfter a quick test firewire worked without realtime and no audio group15:16
scott-workhmmm, interesting15:16
ailoWith realtime and no audio group, jackd would crash15:16
persiaThat's a bug, and ought get sorted.15:16
ailoWe only need to be in audio group to get full performance on Natty15:17
ailoOr, what the drivers and such will allow15:17
persiaBut why?  If JACK is crashing, there's a bug in JACK that ought be fixed.15:17
ailoHaven't confirmed on a another machine. Holstein had a crash when not being in audio group and using realtime15:18
persiaOK.  We need an apport crash report on that bug, and it oughtn't be that hard to fix JACK so it doesn't crash.15:19
ailoOk. Let's tell holstein to check it out.15:20
persiaIf there's some permissions issue, JACK may have to send some sort of warning, but crashing is always wrong.15:20
paultagpersia: it fails with dignity15:20
paultagpersia: the app stays alive and opens a message box suggesting to change permissions15:20
persiaAh, that's not a crash then.15:20
paultagpersia: +115:20
persiaAh, ugh.  It's a limitation due to limits.conf: the choices are 1) set rtprio for @audio, and put users in the "audio" group, or 2) set rtprio high by default.15:26
AutoStaticThat's it. You don't have to be in the audio group to use JACK in realtime15:27
persiaOnly if PAM is configured like that.15:27
AutoStaticBut you do have to be in that group when you want to use FireWire with the JuJu stack15:27
AutoStaticYes, only if PAM is set up for the audio group15:27
AutoStaticrtprio at least, so pam_limits15:27
persiaI don't know enough about PAM, but I suspect there's a way to set rtprio for users logging in at console, and not for remote users or system users, which would seem to better match "Just Works".15:27
AutoStaticYes that's possible15:28
persiaYes, that's how we do it in Ubuntu today, with pam_limits and limits.conf15:28
persiaDo you know how?15:28
AutoStaticyou can set that in the /etc/pam.d directory15:28
persiaDo you know what config needs to be dropped there to make pam_limits set rtprio for the local logins only?15:29
persia(we'd also want to set memlock, but that's a trivial extension once there is an example for rtprio)15:29
AutoStaticJust a sec, I know a bit of PAM but I am no guru either ;)15:30
persiaThanks a lot for looking into this! :)15:30
AutoStaticgrep /etc/pam.d -lr -e "pam_limits"15:32
AutoStaticThat will give you a list of authentication stuff that uses pam_limits15:33
persiaI suspect we want it for gdm/kdm/gdm-autologin at least.  Maybe some others that I'm not seeing immediately.15:33
persiaBut looking (for example) at the gdm script, I'm not sure how to set specific pam_limits config there, only in the case where that is the source of session initiation.15:34
AutoStaticby adding conf=/path/to/limits.conf15:36
AutoStaticOr /path/to/limits.d/gdm.conf for example15:36
AutoStaticIt will override the default settings15:37
persiaSo something like "session required pam_limits.so conf=/etc/security/limits-rtprio.conf" ?15:37
AutoStaticSo if I understand the pam_limits manpage correctly it shouldn't process any other files in /etc/security/limits.d/15:37
AutoStaticyes15:37
* persia reads the manpage again15:39
persiaThe trick is that we *want* it to process the rest of limits.d/ but just add the one special extra config conditionally15:39
AutoStaticAh, oh yeah, well, I'm reredinmg it too15:39
AutoStaticrereading15:40
AutoStatic"If a config file is explicitly specified with a module option then the files in the above directory are not parsed."15:40
persia|If /etc/security/limits.conf has some syntax to specifically include /etc/security/limits.d/, I'd be a lot happier, because then it would be easy.  Unfortunately, it's not clear to me how to make it work.15:40
persiaRight.15:40
AutoStaticAnd the above directory is /etc/security/limits.d15:41
AutoStaticWell, I think if you include the common-* files in /etc/pam.d and put "session required pam_limits conf=/etc/security/limits.d/rtprio.conf" later in the file you should get what you're aiming for15:43
AutoStaticAh, but you want it to process everything in /etc/security/limits.d :$15:44
AutoStaticNot sure how to set that up15:45
persiaForcing conf= seems like it will break stuff for other packages.15:48
persiaWhat's the risk of setting rtprio=99 for everyone?15:48
AutoStaticNot sure, I'm not a security expert15:50
* persia goes to ask the experts15:50
AutoStaticBut it could lead to programs trying to take advantage of this, setting their SCHED_FIO/SCHED_RR prio to 99 and stall the PC maybe?15:51
AutoStaticOr am I talking BS here ;) ?15:51
AutoStaticSCHED_FIFO15:51
persiaNo, but we're making that happen for all our real users anyway: the risk is more about system users.15:51
AutoStaticPAM does allow you to set optional control values15:53
persiaWhat does that mean, in practice?15:53
AutoStaticThat the success or failure of a module doesn't really matter15:55
persiaHow does that help?15:56
AutoStaticUnless it is the only module associated with a certain service15:56
AutoStaticmaybe it is possible to set a session required pam_limits15:57
persiaThat's the default.15:57
AutoStaticbut also a session optional pam_limits conf=/yadiyadi/ya for gdm15:57
persiaMy fear there is that it might not succeed, which would mean that some users might not get the rtprio setting, and it would be hard to troubleshoot for the support team.15:59
AutoStaticYeah, I think you're right15:59
persiaSince the current support team is 80% holstein, I'm *really* not tempted to do anything that would increase the workload :)16:00
persiaSeems diwic is already on it: http://comments.gmane.org/gmane.comp.audio.jackit/2311016:25
falktxhey guys16:43
falktxis it ok for me to build abogani's low-latency kernel for lucid and maverick ?16:44
falktxquadrispro: ping16:44
scott-worklol persia , that should be holstein's tag line..."Holstein:  80% of the support team"16:49
quadrisprofalktx, pong16:49
falktxquadrispro: hey16:50
falktxquadrispro: laditools is ready now, right?16:50
quadrisprofalktx, ready and uploaded16:50
scott-workoh, quadrispro and persia, i saw your discussion on the licensing this morning16:50
quadrisprofalktx, but we won't have gladish until piem uploads new flowcanvas16:50
falktxquadrispro: we dont need the most recent one16:50
falktxquadrispro: flowcanvas v0.6 works fine16:51
quadrisprofalktx, moving to #lad16:51
quadrisprofalktx, back here :)17:08
quadrisprofalktx, but sources in src are all QT4-based17:09
falktxk17:09
falktxand?17:09
quadrisproah no, now I understand17:10
falktxquadrispro: what is the problem about qt?17:10
quadrisprono particular problems17:11
quadrisproand I'm noticing that it's just a frontend, so no modules get added17:12
quadrisprolooks good17:12
falktxquadrispro: check the 'keyboard.py' thing!17:12
quadrisprohowever, we should wait to have ladish in the archives first17:13
falktxquadrispro: yes, of cource17:13
quadrisprofalktx, yes, looking at it right now17:13
falktxquadrispro: my app is not even beta right now...17:13
quadrisprofalktx, ok, I'll stay tuned ;)17:14
falktxquadrispro: the keyboard doesnt act by mouse, only by keys17:14
quadrisprofalktx, why not help this poor guy in writing a small bunch of manpages for ladish? :)17:14
quadrisprohi scott-work persia 17:15
falktxquadrispro: i dont really know how to write manpages...17:15
quadrisprofalktx, it's easy, just a bit tedious :)17:17
falktxquadrispro: btw, i made a (complete) patch for mumble jack support17:17
quadrisprofalktx, please take a look -> http://goo.gl/Dn9R117:18
quadrisprofalktx, ah mumble, I remember we had a chat about it17:19
quadrisprobut please get in touch with the Debian VoIP team17:19
falktxquadrispro: better, I'll send upstream17:20
quadrisproof course the right way17:20
quadrisprobye guys17:41
falktxdamn it I hate natty now18:31
falktxeverything fails to build!!18:32
falktxarrrrg18:32
scott-workwhy are they failing to build?18:48
falktxlink error, like composite did18:49
holsteinpersia scott-work lol18:55
holsteini'll read the scroll-back here in a bit18:56
holsteinabout that test18:56
holsteini got some new mics i want to check out for a bit though :)18:56
persiafalktx, After reading http://liw.fi/manpages/, I've suddenly found it really easy to write manpages.  I highly recommend it.19:33
falktxpersia: nice! thanks19:34
scott-workoooh, a new bookmark :)20:41
ScottLthanks for the persia for the direction and help today23:53
persiaNo problem.  I haven't seen diwic around: I recommend chasing up with him about getting JACK to use rtkit so we don't need the limits.conf hack or the audio group (because it may well fall out of my working set before it happens if I don't catch him soon).23:54

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!