[08:47] morning all [08:52] rye: [08:53] rye: didn't you test the speed throttling resently? Was that on Maverick? [10:32] Good day! [11:23] hello karni [11:23] hi duanedesign ! :) [11:25] karni: do you read Ubuntu Planet? [11:25] duanedesign: occasionaly. I should do that more often. I assume there's been something interesting :)? [11:25] jcastro did a nice blog post on Ubuntu One streaming [11:26] karni: he went on a road trip and used the U1 streaming service for music. overall he had a great experience it sounds like [11:26] duanedesign: that I did read on *his* blog ^ ^ yes! awesome [11:27] what version of Ubuntu are you using? [11:28] another bug report about throttling not working :\ [11:29] i have not set up a Maverick VM yet since i upgraded to Natty [11:29] duanedesign: I'm on 10.04 with broken U1 (long story). about the throttling - there's been an issue about the implementation and how twisted handles I/O [11:30] ahh [11:30] duanedesign: on my last build of the Android app, I implemented it the other way around - it *first* waited as long as it should, ant *then* wrote/read data. :) [11:30] :) [11:30] so their is a master bug for this? [11:31] duanedesign: I'm 99.5% sure there is, yes. [11:31] cool, thanks [11:32] duanedesign: sure, np :) [12:25] hello world! [12:26] \n [12:26] :-) [12:26] :D [13:01] nessita: ahoy ahoy, I apparently got the kdeui for ubuntu-sso working :D [13:01] apachelogger: that's great news! [13:04] nessita: we'd only need some runtime detection whether to use kde or gtk [13:04] that said, the module path is the same, except with kde instead of gtk ^^ ... ui_module='ubuntu_sso.kde.gui', ui_class='UbuntuSSOClientGUI', [13:12] apachelogger: right, so your kde class should query the dbus service with those params [13:12] you kde class -> I meant your kde implementation [13:12] :-) [13:13] nessita: I am thinking about syncdaemon here ... or really any API consumer that would want to have platform integrated appearance [13:13] i.e. make a sane default choice unless an application requests a specific ui [13:14] (actually I do not think an application should ever need to specify a particular ui) [13:15] apachelogger: I'm not sure I follow :-/ [13:19] nessita: let's say I write an application that uses USSO, somewhere in my code I have something like ::: login_or_register_to_get_credentials("myApp", "http://example.com", "halp", 0) ::: what GUI will be used? [13:20] apachelogger: right, that is the old (deprecated) API. The recommended API since natty is: (looking link now) [13:21] https://wiki.ubuntu.com/SingleSignOn/UbuntuSsoClient -> register(String app_name, Dict of {String, String} ui_settings), then: [13:21] Additionally, this 'ui_settings' can provide 2 extra keys to define which UI module and UI class should be used in case the service needs to open a graphical interface to the end user. These keys are: 'ui_class': the name of a class that lives within 'ui_module' and that accepts proper parameters (TODO: document parameters). An example of this class can be seen at ubuntu_sso.gtk.gui.UbuntuSSOClientGUI. 'ui_module': a string pointing to a python module tha [13:24] goo morning ubuntuone! [13:24] nessita: how is that portable? [13:24] hi ralsina [13:25] apachelogger: you can make the dbus call so you use your own client, providing the module path and class name. This way you can provide a kde client or a wx client [13:26] why would I want to do that? [13:27] hey ralsina [13:28] hi fagan [13:28] apachelogger: I'm not sure what you mean. But we designed this service having in mind the following: there is an underlying dbus service which is widget toolkit agnostic, and offers the SSO service. Since the SSO service may need to interact with the end user, when you call login, or register, you need to provide an UI [13:28] apachelogger: on windows we have only a qt ui, for example [13:28] apachelogger: so, you can use this service with "any" UI you want (the UI has to behave in a particular manner) [13:29] im off today because its a holiday here but if you need anything just ping and ill get to it since on thursday I didnt have a lot to do [13:29] nessita: that is what I do not get, why would the API consumer want any UI [13:29] like explicitly [13:29] wouldn't it be much better to make a sane choice *for* the API consumer [13:29] fagan: ok, no problem [13:29] apachelogger: what do you mean by "API" consumer? you as a kde dev or the end user? [13:29] the developer [13:30] ralsina: so if you finished some of the screens I can do up that testing thing for the designers and you can keep moving on them [13:30] apachelogger: well, the UI module and class is optional, and we have a sane default (the GTK UI) [13:30] fagan: nope. Working on it now. Turns out my family wanted to see me on weekends. Who could have imagined that :-( [13:30] that is not the defintion of a sane default [13:30] ralsina: hah [13:30] a sane default is one that actually is sane [13:31] loading GTK libs in a KDE envrionment is not sane :P [13:31] (not unless necessary) [13:32] which is why I think that the logic within credentials ought to be changed around to actually run magic and make it absolutely unnecessary for the API consumer to provide any ui settings at all [13:32] apachelogger: there is no KDE UI yet. So the sane default, right now is still gtk. [13:32] the API consumer would deploy its query, credentials.py looks for installed UIs, checks the envrionment it is executed in, makes a choice on what UI might be best suited [13:33] apachelogger: ah, I now see what you mean. Well, since we don't want to query what desktop are we in, we planned on resolving that by proviging 2 packages, one for the gtk UI and other for the kde UI, so the user has the choice to have the client he wants [13:33] apachelogger: we decided we don't want to make the decision ourselves [13:34] fair enough [13:34] nessita: in that case we'd need a change in credentials.py to allow for this ^^ [13:35] apachelogger: care to be more specific? as far as I know, credentials.py is UI-agnostic enough [13:35] nessita: it defaults to GTK [13:35] apachelogger: right, but is only a string. The caller can override this as much as he needs. [13:35] is not like we're importing gtk modules [13:37] if you want to allow the user to switch the frontends depending on what package they got installed, you ought to change the default module to something like ubuntu_sso.default.gui [13:37] apachelogger: let me see the code, it's been a long time since I've been going thru it [13:38] then have the GTK ui split into a seperate package, link ubuntu_sso/gtk to ubuntu_sso/default and set it to provide ubuntu-sso-gui [13:39] that way there can be a KDE ui, or a Qt ui, or a WX ui, all conflicting, replacing and providing ubuntu-sso-gui and containing a link to ubuntu_sso/default [13:39] apachelogger: but that way you can't use different uis depending on the session you are in. Even if we are not doing that now, we may do that in the future when there is more than one ui. [13:39] by conflicting, replacing and providing I mean the deb package relationships btw :) [13:40] besides, I really need sso not to break before june ;-) [13:40] apachelogger: right, we want to be able to have both kde and gtk clients installed at the same time [13:41] ralsina: yeah, we can't add this as a SRU, it will be work for O [13:41] ralsina: apachelogger: we decided we don't want to make the decision ourselves [13:41] nessita: I mean in trunk, I am working on the SSO qt ui there [13:41] I feel you should have a discussion [13:41] either the design goes one way or the other [13:41] UDS is a great place for this discussion [13:41] apachelogger: why? [13:42] doing one and then the other smells like odd behavior change to me [13:42] apachelogger: what would be "one" and "other"? /me is lost [13:43] apachelogger: we can add the desktop detection any time we want. Changing conflicts in packages is much more messy. [13:43] how is changing conflicts messy? :O [13:43] So, we will probably discuss this a lot on UDS or full team sprint and work on a more clear direction on the O cycle [13:44] apachelogger: because if we adda c onflict, we uninstall the qt ui (for example) and we can't tell you to install it back :-) [13:44] nessita: either you do not make the descision of what UI to use within ubuntu sso (thus only providing a default, explicitly recommending to the API consumer to provide their own UI for integration), or you do [13:45] ralsina: http://www.debian.org/doc/debian-policy/ch-relationships.html#s-virtual [13:45] apachelogger: I disagree. You can have 2 binaries, ubuntu-ss-login-gtk and ubuntu-sso-login-kde, like usb-creator [13:45] and in this case, by binaries I mean services [13:46] what is the point of having a gui agnostic service then? [13:49] apachelogger: we have a layered design. The GUI-agnostic service is the one that knows all the SSO logic and that unifies all the calls to the SSO server. The UI-specific service, would be the one that knows how to call the agnostic service with a specific UI. So the client apps can decide what UI-specific service use. For example, the U1 gtk control panel, will call the gtk sso service, and a KDE control panel will call the kde sso client [13:51] apachelogger: what I mean is, the UI agnostic service is useful for devs like you. The UI specific service is useful for 3rd party apps that require desktop SSO auth [13:52] nessita dobey thisfred alecu standup in 9' [13:52] ack [13:52] oh! right [13:54] nessita: so you want internal decision making on what UI to use [13:54] there is absolutely no reason why a GTK CP would enforce a GTK SSO GUI [13:56] there is however for, who knows, a 3rd party skype client that does some magic file sharing via u1 and wants to provide an ultimate branding experience, in which case they would provide their own gui [13:56] but those are 2 different use cases and one does not affect the other at all [13:56] apachelogger: I'm pausing the chat since I need to write my stand up :-) [13:57] kk :) [13:57] hmm [14:00] me [14:00] hello! [14:01] morning alecu [14:01] me [14:01] alecu, dobey, thisfred, CardinalFang, ralsina? [14:01] DONE: looked into and gave up on bug #762722 (for now) TODO: blueprints BLOCKED: no NEXT: nessita [14:01] me [14:02] Launchpad bug 762722 in ubuntuone-control-panel (and 1 other project) "Opening the Control Panel during sync removes progress indicator (affects: 1) (heat: 6)" [High,Triaged] https://launchpad.net/bugs/762722 [14:02] me [14:02] ralsina: are vds and CardinalFang coming back to dekstop+ or not yet? [14:02] vds: not yet, CardinalFang: soonish [14:03] nessita? [14:03] DONE: Easter holidays + a messy (but fun!) moving between homes. [14:03] TODO: catch up with emails after a 4-day weekend, catch up with bug reports, check that no fire needs to be put away, follow on bug #709494, chat with apachelogger re: ussoc multi-UI-design [14:03] BLOCKED: I'm living in a chaos :-), I may take a day off this Friday to try to put my home in order before traveling to Budapest. [14:03] NEXT: ralsina [14:03] Launchpad bug 709494 in ubuntu-sso-client (Ubuntu Natty) (and 5 other projects) "[SRU] Missing user's name field (affects: 2) (dups: 1) (heat: 22)" [High,Fix released] https://launchpad.net/bugs/709494 [14:03] DONE: nothing (religious holidays!), TODO: SSO screens BLOCKED: not at all [14:03] alecu! [14:04] DONE: thu and fri: holy-days. [14:04] TODO: catch up with mail, review roadmap, make blueprints [14:04] BLOCKED: no [14:04] ooooooohhh blueprints. I have to do those too ;-) [14:04] * nessita too, but she needs some insipiration [14:05] ralsina: alecu nessita dobey: should we have a blueprints planning mini meeting? So we can make sure we don't all do the same ones, or more likely skip half of them? ;) [14:05] oh, me [14:05] thisfred: we can have that tomorrow, because mandel is off today [14:05] λ DONE: beyond-beyond-last-minute urgent fix upload, discussion, nightlies almost back to proper building, started on bug #768911 [14:05] Launchpad bug 768911 in banshee (Ubuntu) (and 1 other project) "Natty (beta 2) Banshee Geo IP Detection Amazon Defaults to U.S. (affects: 1) (heat: 6)" [Undecided,Invalid] https://launchpad.net/bugs/768911 [14:05] dobey, go! [14:05] λ TODO: finish #768911, finish nightlies fixes, maverick/lucid SRUs [14:05] λ BLCK: None. [14:05] ralsina: sounds good [14:06] any closing comments? [14:06] thisfred: i'll make it easy for you and just won't do any :) [14:07] hmmm be nice to each other, eat vegetables? :-) [14:07] eom! [14:07] mmmmm vegtetables [14:07] mmm bacon [14:08] * fagan resisted the temptation to participate in the standup and say played some call of duty todo read a book..etc [14:09] bacon should be a honorary vegetable [14:09] I don't see why not, when ketchup is [14:09] ralsina: pigs are vegetables. have you ever tried talking to one? they just sit there in the mud. [14:09] that wacky old Reagan [14:10] tomatos are only kinda vegetables [14:10] I make home-made ketchup and use nothing but vegetables (and vinegar and sugar, which are vegetable products) ;-) [14:10] fagan: kinda like strawberries [14:10] mmmmmmmmmmmmmmmm strawberries [14:10] fagan: tomatoes are fruits [14:10] ralsina: in the 80s in the US when schools were required to serve vegetables in lunches, Reagan decided that ketchup would count as a vegetable... [14:11] ralsina: but they taste like vegetables [14:11] sugar isn't a vegetable product [14:11] dobey: well it is from sugar cane [14:11] dobey: it's made out of sugar cane [14:11] dobey: beet sugar us [14:11] is [14:11] high fructose corn syrup is [14:11] oh then all sugar is vegetables that justifies me using so much sugar in my coffee [14:12] sugarcane is a grass [14:12] transfat is a vegetable [14:12] dobey: ooooh [14:12] grasses are vegetables. [14:13] palm trees are grasses too, yet palm heart is a vegetable [14:13] there are no real hard scientific definitions of either vegetable or fruit [14:13] well, not mineral, not animal, not fungus? :-) [14:13] much like toad vs. frog, incidentally [14:14] uhm [14:17] anyway [14:17] how do i make django.test.client.Client use a a specific IP as the REMOTE_ADDR? [14:18] nessita, ping [14:18] nessita, chipaca mentions "the user testing report by mpt" in a mail from last thursday... do we have a detailed report on that? [14:18] alecu: pong [14:19] alecu: isn't that what lucio sent to our mailing list? [14:21] fagan, and corn is a grain, not veg. Sorry. [14:21] nessita, yes, I found it. I remember it said it was "the summary", so I was wondering if you guys were referring to some other more detailed report on u1. [14:22] alecu: ask charline for specifics [14:23] alecu: I don't have any specifics, maybe Chipaca or cparrino have them? or, what dobey says [14:23] alecu: the full report isn't written yet afaik [14:23] alecu: mpt just sent the summary to ubuntu-devel [14:23] cool [14:23] in the big unity thread [14:24] right. after which it promptly got /.ed [14:25] * alecu should join ubuntu-devel. [14:25] thanks all! [14:29] nessita: the dbus stuff in /credentials is deprecated? [14:30] apachelogger: yes sir (I'm finishing with some email to fully restart our conversation, sorry for the delay) [14:31] alecu: we were all told to do that by our fearless leader like a month ago ;-) [14:33] nessita: https://bugs.launchpad.net/ubuntu/+source/ubuntu-sso-client/+bug/770269 :) [14:33] Launchpad bug 770269 in ubuntu-sso-client (Ubuntu) "deprecated DBUS interfaces are not marked as deprecated (affects: 1) (heat: 6)" [Undecided,New] [14:33] ralsina, which one is the fearless one? [14:34] alecu: chipaca [14:34] alecu: can we somehow mark a dbus API as deprecated? [14:35] nessita, I don't know of any other way to mark them as deprecated, other than doing so in the docs. [14:36] nessita, I'm checking anyway. [14:36] apachelogger: how would you suggest we mark the /credentials iface as deprecated? [14:37] nessita: I think you will have to annotate each method within as deprecated [14:37] nessita, alecu: there is no way, no. other than to respond with an error of "This is deprecated" over D-Bus, but that's not really helpful [14:37] * apachelogger has no idea how to do that in python ^^ [14:37] dobey: right, specially since we don't want to break that API until natty [14:37] apachelogger: i think that just marks it in the docs though [14:38] * alecu is looking how to do that in python. [14:38] alecu: doc strings. [14:38] apachelogger: I'll see if I can add a deprecated warning to the code itself, but the dbus caller may never get to see that. The doc does mark that iface as deprecated though :-) [14:39] nessita, dobey: here's the link that apachelogger pasted: http://dbus.freedesktop.org/doc/dbus-specification.html#introspection-format [14:39] dobey: as long as the introspection reflects deprecation that ought to be sufficient IMHO (for natty anyway) [14:39] ok, I'll prepare a branch for trunk then :-) [14:39] apachelogger: thanks for the pointer! [14:39] 'Method, interface, property, and signal elements may have "annotations"....''' -> org.freedesktop.DBus.Deprecated = true [14:39] yw [14:40] oh [14:42] nessita, what will you put in that branch? [14:46] nessita: we can swap the weekend between summit and uds? Yay! ;-) [14:46] * ralsina really needs that manual [14:47] ralsina: that's what I've asked you before... :-) [14:47] ralsina: and you said yes! [14:47] hmmm... you didn't mention WHAT day you were swapping. But yeah, I think it's ok. I approved it anyway [14:48] I thought you were going to swap the travel day [14:49] ralsina: I'm travel on a weekday, so I don't think we can swap that [14:49] oh, well, I'll ask those in the know later. [14:50] * ralsina goes away for 2 days and ends with 1200 emails. Cute. [14:51] nessita, dobey, apachelogger: there's no support at all in /usr/lib/pymodules/python2.7/dbus/service.py to add any "annotation" tags in the introspection xml. That means no "deprecated" tags for natty. [14:52] alecu: we really need to move away from dbus-python anyway [14:52] alecu: I was planning on updating the docstrings, at least, since apparently apachelogger used that as documentation for building his app [14:52] alecu: and for 11.04 we aren't doing anything else now anyway unless it's a bug fix. [14:53] 11.04 is done. [14:53] dobey, right, and no srus for this. [14:54] i have no idea how to use GDBus in Python though [14:55] dobey, why do we have to move away from dbus-python? is there anything else broken? [14:56] alecu: well, it is unmaintained, deprecated, and will no longer be updated [14:57] dobey, deprecated? do you mean the C library below it, or dbus-python itself? [14:57] alecu: it is replaced by GDBus, which is in libgio [14:57] alecu: both [14:58] alecu: pygtk/pygobject are also finished. we should move to using libraries via gobject-introspection [15:00] dobey: can you please make a master bug affecting u1cp, ussoc, and any other UI you can think of? [15:02] so, we now start to migrate everthing to C-written-in-python, or you guys want to contribute to the annotations to make it pythonic? [15:02] nessita: probably a blueprint with lots of bugs, because we also need to do other stuff, like use GTK+ 3.x [15:02] ralsina: i don't understand that [15:03] dobey: wanna take over that? [15:03] dobey: the raw instrospection-based bindings usually suck and do things like returning error codes [15:04] ralsina: they aren't bindings, and i haven't really tried using it from Python, but yes I know there are lots of issues with the gir bits, as i've tried to use some from vala [15:05] ralsina: if stuff is broken we just have to file bugs, and wait for fixes or work around or fix it ourselves, sadly :( [15:05] we're also doing some rather nasty things in some of our UI [15:05] dobey: exactly. There is a mechanism to improve them for python (and other languages) by adding annotations, IIRC [15:06] nessita: https://lists.ubuntu.com/archives/kubuntu-devel/2011-April/005145.html [15:07] ralsina: depends. using from vala has more problems because it's a compiled language though, so it's not necessarily that method annotations are wrong, but there are higher level issues with the girs themselvs :( [15:08] dobey: :-( [15:08] and don't even get me started on webkit :) [15:09] * fagan thought webkit was good [15:09] fagan: webkit *is* good [15:09] haha [15:10] fagan: at least it's better than the alternatives. Specially now that there is none ;) [15:10] it's really good at causing me mass frustration [15:10] apachelogger: so, I can restart our conversation now. Let me read the last thing you said before stand up. [15:10] ralsina: sounds good anyway :) [15:10] and its the busiest upstream id say other than the kernel [15:11] apachelogger: I disagree with "there is absolutely no reason why a GTK CP would enforce a GTK SSO GUI", I think that a given UI has the right (obligation?) to choose the proper sso UI to match the look and feel [15:13] nessita: sure, if there is an explicit look and feel to match, such as theming of 3rd party applications [15:14] "system native" applications OTOH should strive for the greatest possible overall integration as their is no perceived look and feel other than the one of the environment [15:14] apachelogger: I'm not sure what you mean by "syste native" apps. Care to give an example? [15:14] say a user is using the GTK CP in a KDE envrionment for whatever reason there might be ... chances are the user will not know what GTK is and why it looks different [15:15] lets further assume that the GTK theme on the setup is not matching whatever the KDE envrionment is using [15:15] apachelogger: right, but in that scenario makes sense to present the user the GTK sso screen, for consistency sake [15:15] so the GTK CP is introducing an alien look and feel [15:15] applying that to the SSO as well does not make it a more wholesome look and feel [15:15] it only makes more things appear alien [15:16] nessita: 2 times bad consistency still doesnt make good consistency though :D [15:17] apachelogger: jeje. But the SSO screen is (ideally) made to be transient for the caller window, so from my POV is better to show a consistent look and feel between the CP and the SSO then showing those 2 as alien between eachother. But I know this is a subjective opinion, and others my agree with you [15:18] nessita: ultimately I believe you should go ask the design team [15:18] they surely will know best anyway :) [15:18] apachelogger: I agree [15:21] apachelogger: I will make a note to talk with mpt or someone else from the design team. Are you attending to UDS? [15:22] yep [15:22] apachelogger: please ping me about this [15:22] aye [15:23] * alecu needs to be away for 30 minutes [15:23] over and out. [15:29] * nessita reboots after update [15:49] thisfred: ping [15:49] nessita: pong [15:49] thisfred: you aware of bug #768576? [15:49] Launchpad bug 768576 in desktopcouch (Ubuntu) "desktopcouch keeps becoming zombie (affects: 1) (heat: 6)" [Undecided,New] https://launchpad.net/bugs/768576 [15:50] nessita: not of this specific one no, but we have seen one of the d-c processes turn zombie before. I thought chad worked on that during the sprint though. [15:50] nessita: this user seems to have problems with multiple unrelated processes though [15:51] thisfred: can you confirm is this is the known issue? [15:51] (please) [15:51] yeah I'll look if there is a master bug [15:52] thanks! [15:52] dobey: when you have a spare moment, would you please share your thoughts on bug #768738? [15:52] Launchpad bug 768738 in ubuntuone-client (Ubuntu) "nautilus crashed with SIGSEGV in g_str_hash() (affects: 1) (heat: 6)" [Undecided,New] https://launchpad.net/bugs/768738 [15:56] dobey: and also, does bug #769073 makes sense to you? [15:56] Launchpad bug 769073 in ubuntuone-client (Ubuntu) "package libsyncdaemon-1.0-1 1.6.1-0ubuntu3 failed to install/upgrade: ошибка записи в «<стандартный вывод>»: Победа (affects: 1) (dups: 3) (heat: 24)" [Undecided,New] https://launchpad.net/bugs/769073 [15:56] (error message was translated by rye) [16:00] no, it looks like a fluke on that user's system. maybe something happened in dpkg itself, but doesn't seem like an ubuntuone issue [16:01] CardinalFang__: we never did diagnose the the zombie desktopcouch-service process, right? I still see it. [16:02] thisfred, I found a problem with the parent killing children processes, but not a child dying a the parent not being there to wait() and let the kernel remove that zombie. [16:03] nessita: and the first is a dup as per the one comment on it, i believe [16:04] nessita: and in bug #680968 it makes no sense to me why the code is even trying to deal with non-local paths [16:04] Launchpad bug 680968 in ubuntuone-client (Ubuntu) "nautilus crashed with SIGSEGV in g_str_hash() [ubuntuone_nautilus_observed_file_unref] (affects: 16) (dups: 14) (heat: 69)" [Medium,Triaged] https://launchpad.net/bugs/680968 [16:04] CardinalFang__: for me it's pretty reliable: boot into natty, and after a minute or so I see 3 d-c-service processes, one of which zombified [16:05] thisfred, weeeeeeird. Not here. [16:05] dobey had the same I believe? [16:06] dobey: right, the last bug you linked has several dupes, so seems like somehow we're messin that up :-/ [16:06] messing* [16:06] there is a known problem with wait() and zombie children in python, which I can't really remember [16:06] nessita: yeah, so i guess we need to fix it soon [16:07] CardinalFang__: i'd seen that before as well, yes [16:08] One reason for those was that the processes are writing something to stdout/stderr and noone was reading that. [16:08] at least that's the most common reason for python zombies when using popen or subprocess [16:08] that sounds like it may be a good place to start looking [16:09] anyway, this guy is claiming it's making his whole system freeze, which I think we can safely say is probably not the case [16:09] yeah, zombies don't do that [16:09] ralsina: this is using fork() [16:10] dobey: hmmmm that's trickier. [16:10] thisfred: yeah, if my system froze because of desktopcouch, yall would be hearing about it fo sho. [16:10] dobey: I know, you're our canary :P [16:10] ralsina: yeah, subprocess issues would be easy peasy :) [16:11] anyway, it's the same principle. A zombie is there because it's dead and the parent has not called wait() yet. [16:12] should be fixable by handling SIGCHLD [16:13] or setting SA_NOCLDWAIT [16:15] CardinalFang__: bug #768576 assigned to you. I couldn't find another master bug about the zombie processes, so this is it from now on. [16:15] Launchpad bug 768576 in desktopcouch (Ubuntu) "desktopcouch keeps becoming zombie (affects: 1) (heat: 6)" [Medium,Confirmed] https://launchpad.net/bugs/768576 [16:20] thisfred, ah, I expect a SRU on day 1 of Natty. Fun. [16:20] CardinalFang__: I'm looking at the code now, maybe I can figure it out myself, but I don't hold super great hopes [16:21] thisfred, let me push up this branch I have for child processes. [16:27] lunch, bbiab [16:28] lunch! what a great idea [16:28] thisfred, https://code.edge.launchpad.net/~cmiller/desktopcouch/service-exits-properly/+merge/58964 [16:29] thisfred, not about a child exiting. Only when parent is exiting. Doesn't address Zombie state at all, I think. [16:31] right, good to have though [16:31] CardinalFang__: my suspicion about the zombie child is that it's the migration one, (that's the only one that's supposed to stop at some point right?) [16:32] Not sure why it would zombie rather than exit in some cases and not others. [16:33] thisfred, yes, it should sleep a minute, do something, and then exit. [16:35] CardinalFang__: zoombies happen after they exit ;-) [16:35] CardinalFang__: do you need to know the exit code of the child? [16:36] ralsina, right, I'm not overriding SIGCHLD or anything. I don't need the value, no. [16:36] then use sigaction and set SA_NOCLDWAIT [16:37] that way zombies are impossible. Or supposed to be impossible. [16:38] "SA_NOCLDWAIT [16:38] If set, and sig equals SIGCHLD, child processes of the calling processes shall not be transformed into zombie processes when they terminate." [16:41] or set SIGCHLD to handler SIG_IGN and the same thing should happen [16:42] by default it should be SIG_IGN. Oh, well. [16:43] ralsina, Yeah, that's what bothers me. I don't set the handler to anything, but I am in a gobject mainloop, so there may be some magic I don't know about. [16:43] You can check what the handler is [16:44] maybe someone else is setting it, yes [16:47] I have problems with evo contacts in desktop couch [16:47] I have 2600 contact in ubuntuone [16:48] they have been synchronized to desktopchouch [16:48] Evo shows a timeout [16:50] tuhl: so they show up in Couch but not Evolution? [16:50] duanedesign: yes [16:50] I see them all in the web browser [16:52] tuhl: can you try the steps under Killing and restarting Desktopcouch http://www.freedesktop.org/wiki/Specifications/desktopcouch/Documentation/Troubleshooting [16:59] duanedesign: I restarted couchdb [16:59] still same problems [17:00] is there anybody using desktopcouch with more than 2500 entries? [17:00] ok [17:00] is the evo adress book loading all contacts? [17:00] into memory? [17:01] can you try: evolution --force-shutdown; /usr/lib/evolution/e-addressbook-factory [17:01] that might print a clue to the issue in the temrinal [17:02] tuhl: oh, what version of ubuntu? === CardinalFang__ is now known as CardinalFang [17:16] latest 11.04 beta [17:16] all updates [17:18] termintion of evo adr book and restart of evo does not help [17:18] did any errors get printed to the terminal? [17:19] if you run: /usr/lib/evolution/e-addressbook-factory [17:19] and then open Evolution and click on the contacts [17:22] ** (e-addressbook-factory:5819): DEBUG: Creating cache at /home/tuhl/.cache/evolution/addressbook/couchdb___127.0.0.1/http://127.0.0.1:58057/cache.xml [17:23] (e-addressbook-factory:5819): Json-CRITICAL **: json_object_get_string_member: assertion `node != NULL' failed [17:24] (e-addressbook-factory:5819): Json-CRITICAL **: json_object_get_string_member: assertion `node != NULL' failed [17:24] book_view file uref [17:24] tuhl: coulf [17:24] could you use paste.ubuntu.com [17:25] I know it wasnt much but pasting tracebacks in irc is ugly === beuno is now known as beuno-lunch [17:28] http://paste.ubuntu.com/598789/ [17:28] the XML file which is generated in .cache contains all vcards [17:30] dobey: when you come back from lunch, would you help me with bug #765441? I know we ship DEBUG by default in nightlies, so I was wondering if we somehow leaked this setting to the official release [17:30] Launchpad bug 765441 in ubuntuone-client "logging defaults to DEBUG in released version (affects: 1) (heat: 6)" [High,Confirmed] https://launchpad.net/bugs/765441 [17:31] the evo couchdb backend seems not to be very robust [17:31] why does it generate a seperated XML file? [17:31] it should query the couchdb directly [17:34] I even would call it unsuable [17:35] nessita: hrmm, it should be automated but maybe someone broke it. i'll look at it [17:35] dobey: thanks! [17:44] tuhl: sorry for the delay. Been looking through bug reports. bug 727370 sounds similar to your issue. Probablly be best if you talked with rodrigo. I think he is on holiday today. [17:44] Launchpad bug 727370 in evolution-couchdb (Ubuntu) (and 1 other project) "Contacts not syncing with my computer (affects: 27) (dups: 16) (heat: 256)" [Low,Confirmed] https://launchpad.net/bugs/727370 [17:44] rye: might know something about it but i have not seen hime today either [17:47] thisfred, of that patch, fixed up test to include it and attached bug. [18:03] duanedesign: rodrig is the core developer of the evo couch backend? [18:05] alecu: hey there, remember that priority High is used for important bugs affecting a lot of users. The scroll bar report, though it may be annoying for some users, is not a bug as in a breakage or inability to use the app, so I'm lowering the priority [18:05] nessita, great [18:07] tuhl: i know he has been wotking on it lately. If possible can you come back by tomorrow around this time? I am sure you can catch him then and we can get you more information about your bug. [18:07] hello alecu [18:08] hi duanedesign [18:08]