=== chihchun_afk is now known as chihchun
=== chihchun is now known as chihchun_afk
=== chihchun_afk is now known as chihchun
=== chihchun is now known as chihchun_afk
=== chihchun_afk is now known as chihchun
=== chihchun is now known as chihchun_afk
AceLanHi, is it possible to run camera-app on desktop environment? http://paste.ubuntu.com/9210403/ # This is what I got to run camera-app from desktop07:54
dholbachgood morning08:20
rpadovanioSoMoN, o/ As you see, I didn't fix tests for my branch, I'm sorry - Also I don't think I have time this and next week. Do you want I push on a shared branch, so you can fix tests, or do you prefer to wait and I fix them as soon as I have some free time?08:43
tsdgeosMirv: i could not reproduce any restart08:46
tsdgeosdo you have any crash log? is there still any ppa with those dbus changes in?08:47
oSoMoNrpadovani, there’s no rush at this point, and I’m busy with other things, so take your time :)08:51
rpadovanioSoMoN, ok cool :-)08:52
Mirvtsdgeos: are you sure you used the PPA's version, since as of Friday the same version number is in archives? let me do a ubuntu6 to the 027 silo09:03
Mirvtsdgeos: the upstart logs I provided already, I've some .crash files but I cannot be sure if they were during testing that PPA or during my other PPA autopilot testing: http://people.canonical.com/~tjyrinki/qt532/possible-dbus/09:05
tsdgeosdpm: any idea of what may be wrong in https://bugs.launchpad.net/ubuntu/+source/indicator-power/+bug/1395640  ?09:21
ubot5Launchpad bug 1395640 in indicator-power (Ubuntu) ""Battery Critical" is not translated/translatable" [Undecided,New]09:21
tsdgeosMirv: no, not totally sure09:21
dpmtsdgeos, no, I don't know, I actually think I reported it a while ago too09:23
dpmtsdgeos, I can't find my bug, perhaps I didn't report it in the end, but I had a conversation with Mirco about it. Could it be that the wrong domain is set?09:25
tsdgeosdon't think so09:25
tsdgeosseb128: ↑↑ do you know how this works?09:33
seb128tsdgeos, dpm, I think it's https://code.launchpad.net/~seb128/indicator-power/update-translations-list/+merge/238258 ?09:35
seb128tsdgeos, dpm, bug #139170209:35
ubot5bug 1391702 in indicator-power (Ubuntu RTM) "Low battery notification not localized" [High,Triaged] https://launchpad.net/bugs/139170209:35
mardydednick: hi! Sorry for Friday, I got disconnected; let me know when we can resume the discussion (bug 1395028)09:35
ubot5bug 1395028 in QtMir "Dead processes are still shown in a trust session" [Undecided,New] https://launchpad.net/bugs/139502809:35
dpmah, thanks seb12809:36
tsdgeosseb128: this is very confuing09:36
tsdgeosmake .pot creates a correct .pot file09:36
seb128tsdgeos, ?09:36
dednickmardy: no worries. whenever you want09:36
tsdgeosseb128: but something else is using po/POTFILES.in ?09:37
seb128tsdgeos, no, it's just that the fix landed in vivid/trunk09:37
tsdgeosdo we have two scripts/lists in the same project to generate .pot files?09:37
seb128but not in rtm09:37
seb128where commit/branch do you run make pot on?09:37
seb128where do you see the issue?09:37
tsdgeosseb128: i'm in rtm packages09:37
tsdgeosapt-get source09:38
tsdgeosand make pot gave me the correct .pot09:38
tsdgeosbut indeed that file is missing from POTFILES.in09:38
seb128what command is make pot calling?09:38
mardydednick: so, the main usecase is when A is a client application, B is an unconfined process (the Online Accounts interface), and C is another confined process (an Online Accounts plugin)09:39
tsdgeosseb128: http://paste.ubuntu.com/9211624/09:39
seb128ok, yeah09:40
seb128tsdgeos, that project uses cmake, so yeah, the ubuntu build and make pot run different command09:40
tsdgeosthat's horrible09:40
seb128the Ubuntu build/dh-translations integration uses intltool-update09:40
seb128which relies on the POTFILES.in09:40
seb128the cmakery does its own thing09:40
seb128tsdgeos, we should probably commit something like that https://code.launchpad.net/~seb128/indicator-session/update-translation-template/+merge/240458 and delete the POTFILES.in09:41
seb128tsdgeos, that would make it update the pot with "make pot" rather than calling intltool-update09:41
tsdgeoshaving 1 place to do things is better09:41
seb128tsdgeos, what is?09:41
tsdgeoshaving 2 leads to confused people like me :D09:41
seb128cf what I just wrote09:42
seb128I'm going to do a mp for that09:42
dednickmardy: ok. i need to know exactly how the processes are dealing with prompt sessions. ie which process creates the prompt sessions.09:42
tsdgeosseb128: thing is i don't know enough of the debian/ubuntu side to give you the +1 of the .pot file ending in the correct place with the correct name the debian side needs09:43
seb128tsdgeos, dpm, meanwhile we should land https://code.launchpad.net/~seb128/indicator-power/update-translations-list/+merge/241617 for ota if we can09:43
dednickmardy: you said there were 2 prompt sessions?09:43
tsdgeosseb128: +109:43
seb128tsdgeos, there is no "debian side", launchpad imports any .pot it finds in the builddir09:43
seb128but anyway, let it to me, I know what to do09:43
tsdgeosah that's good :D09:43
mardydednick: no, it's only one session; there is another non-UI process which creates the session and starts both processes B and C inside it09:45
mandelogra_, when ever you have the time, can you build android tools with this patch => http://paste.ubuntu.com/9211691/09:45
mandelogra_, is shorter that the previous one and AFAIK it should work09:45
dednickmardy: can you point me to the code?09:45
ogra_mandel, oh, i thought you would just call dbus-send and hand over the address to this09:47
mardydednick: http://bazaar.launchpad.net/~mardy/ubuntu-system-settings-online-accounts/lp1380914-lite/files/head:/online-accounts-service/09:47
mardydednick: mir-helper and ui-proxy09:48
mandelogra_, no need AFAIK, I found the security setting and a workaround :P09:48
ogra_ah, cool09:48
mandelogra_, I'd like to test it, afaik it should work09:48
JamesTaitGood morning all; happy Monday, and happy Celebrate Your Unique Talent Day! :-D09:48
mandelogra_, if we fake the userid the security thinks we are the good guy :)09:48
mandelogra_, and we can call it etc.. if that is the case we can set the env var (so that we only look for it once and so that we can set it for testing) and do all the stuff09:49
dednickmardy: where is the second one coming from?09:49
mardydednick: same code: see that in line 204 and followings, we resue the same trust session if the initiator PID is the same: http://bazaar.launchpad.net/~mardy/ubuntu-system-settings-online-accounts/lp1380914-lite/view/head:/online-accounts-service/mir-helper.cpp#L20409:51
dednickmardy: where is the "re-use"? i seen it calling createPromptSession in setupPromptSession.09:53
mandelogra_, sorry, I found a meme leak, fixed here => http://paste.ubuntu.com/9211814/09:54
* mandel grabs coffee to stop making mistakes09:54
dednickmardy: oh, sorry, it's private vs public methods09:54
dednickmardy: is this easily reproducable on the phone image now? or need tht branch?09:58
dednicknot really sure why it's not working. perhaps a unity8 upstart log might shed some light though09:59
mardydednick: you need that branch, unfortunately. If you want, I could modify your trusted prompt examples (where were they?), it should be rather quick10:00
dednickmardy: lp:~nick-dedekind/+junk/trusted_sessions_app10:01
dednickmardy: not really sure why it's different though. maybe adding fd's twice is buggy or something. i think i only tested using multiple fds at once.10:02
mardydednick: I remember that we agreed that there would be a limit on the number of trusted sessions, but not on the number of processes participating in it10:04
dednickmardy: right. can only have one trust session started by a single process, but there shouldnt be a problem nesting them or having multiple participants.10:05
dednicki mean one trust session per process.10:06
Mirvsil2100: thanks for the quick appmenu qt 5.4 branch! any idea about the problem I still saw with it in the PPA?10:56
Mirvsil2100: I must say I didn't yet upgrade my desktop to the qt 5.4 beta, but I'm doing that now..10:58
sil2100Mirv: hey! Let me take a look at that ;)10:58
* sil2100 the same10:58
Mirvsil2100: ok, and of course there are more important stuff to do, but I'm happy to give you some coding fun :)10:58
MirvI hope some of the bug I've seen so far go away with the release candidate, but that of course does not apply to changed API:s etc11:01
ogra_mandel, ervices.c: In function ‘is_phone_locked’:11:15
ogra_services.c:385:5: error: unknown type name ‘GDBusConnection’11:15
ogra_     GDBusConnection *connection = NULL;11:15
ogra_mandel, i assume the dbus header is missing ?11:16
mandelogra_, uh.. let me double check, one sec11:16
mandelogra_, #include <gio/gio.h>11:16
mandelogra_, sorry11:16
=== _salem is now known as salem_
ogra_and another one ...11:25
ogra_services.c:473:5: error: expected ‘;’ before ‘setuid’11:25
ogra_     setuid(0);11:25
ogra_mandel, ...11:35
ogra_services.o: In function `is_phone_locked':11:35
ogra_/home/phablet/android-tools-4.2.2+git20130218/core/adbd/services.c:426: undefined reference to `g_dbus_connection_new_for_address_sync'11:35
mandelogra_, wtf? I compiled al this correctly...11:35
mandelogra_, `g_dbus_connection_new_for_address_sync' in in gio.h11:36
mandelogra_, let me try again11:37
ogra_do i perhaps need other includes in the makefile too ?11:37
mandelogra_, I keep getting a bloody dpkg-source: info: the patch has fuzz which is not allowed, or is malformed when building every now and then :-/11:41
mandelogra_, might need to point to the gir11:41
mandelogra_, I need to swear.. puto quilt11:44
mandelogra_, forgot to do a add debian/control11:44
ogra_dont do that in the quilt patch11:44
ogra_(dont do the makefile there either)11:44
ogra_that will make quilt a lot more cooperative ;)11:44
mandelogra_, really? ahg...11:45
mandelogra_, this has all changes => http://paste.ubuntu.com/9213217/11:47
mandelogra_, make files etc..11:47
popeywhere did "--list-channels" go in ubuntu-device-flash!?11:48
ogra_mandel, no, it hasnt ... but thats fine :P11:48
popeyaha, under query11:48
* ogra_ assumes you also wanted "#include <gio/gio.h>" in the code 11:48
mandelogra_, one sec, the dependency in coontrol is wrong11:48
mandelogra_, true11:48
mandelogra_, give me another try..11:48
* mandel feels stupid11:49
popeyogra_: any idea what I'm doing wrong here? http://paste.ubuntu.com/9213233/11:49
mandelogra_, I think is ok now = http://paste.ubuntu.com/9213248/11:50
ogra_builds with just the makefile change added (i assume the package already came in as a dep but will add it to debian/control anyway)11:52
mandelogra_, sweet, is better to be explicit.. at least that is what I've learned ;)11:55
mandelogra_, but I really really hate how quilt works in that package11:55
sil2100Mirv: hah, I see the problem with my branch, damn, I feel so stupid now ;)11:58
ogra_mandel, http://people.canonical.com/~ogra/ubuntu-touch/android-tools-adbd_4.2.2+git20130218-3ubuntu37_armhf.deb and the debdiff is http://paste.ubuntu.com/9213359/12:00
mandelogra_, testing very eager to get this out of my plate ;)12:01
ogra_popey, you have to ask sergiusens_ about --list-channels, not sure where it went ...12:05
ogra_popey, try devel instead of 14.09 ... and also give the touch arg12:06
ogra_popey, heh, thats flo ... there was never any rtm promotion for any of the tablets12:07
popeythats why it's old and crusty then12:07
* popey might put proposed on for the lolz12:07
ogra_well, there is nothing in it12:07
popeythanks ogra_12:09
=== sergiusens_ is now known as sergiusens
sergiusensogra_: hello12:11
sergiusensogra_: you have a tail too :-P12:11
ogra_sergiusens, works :)12:11
ogra_minne is natiurally grown ;)12:11
ogra_(together with the goat feet and horns)12:12
Mirvsil2100: ok, problem found is half of problem solve, sounds good! :)12:13
* sil2100 thinks how to test that now12:15
mandelogra_, did you get my last message?12:16
ogra_mandel> ogra_, testing very eager to get this out of my plate ;)12:17
ogra_thats the last i see from you12:17
mandelogra_, ok, so xchat went nuts when I changed wifi12:19
mandelogra_, the package works as expected, there is only one thing I don't like, the client side says 'error: closed' when the screen is locked12:19
mandelogra_, I asked rsalveti to take it for a spin too12:19
ogra_not much we can do about that i guess12:19
=== MacSlow is now known as MacSlow|lunch
mandelogra_, not really, we are lucky that it happens because the phone is closed lol12:21
mandelogra_, would be great if it said 'device closed' lol12:21
ogra_sergiusens, did davmor2 point you to bug 139568212:23
ubot5bug 1395682 in upower (Ubuntu) "Ciborium is disconnecting over night and looping causing multiple messages" [Undecided,Incomplete] https://launchpad.net/bugs/139568212:23
ogra_looks to me like a upower or MMC driver thing12:23
sergiusensogra_: it's incomplete for a reason, but it hell ain't ciborium :-)12:23
davmor2sergiusens: I don't care I'm still blaming you :P12:24
sergiusensand I'm not fixing it12:24
davmor2sergiusens: did you like the screenshot :)12:24
sergiusensor won't try too unless you collect all that info I asked for12:25
sergiusenswill take you a week at least :-P12:25
sergiusensdavmor2: yeah, the screenshot is useless to me12:25
davmor2sergiusens: you're so predictable ;)  I doubt some how that it will only take a week :)12:27
=== dandrader is now known as dandrader|afk
sergiusensdavmor2: you have to read the manpages for all the commands ;-)12:31
davmor2sergiusens: man ciborium: No manual entry for ciborium well that was quick ;)12:32
sergiusensdavmor2: not ciborium, udevadm and such ;-)12:33
davmor2erm no12:33
mardydednick: hi :-) So, I see that your example code works, with multiple apps in the same session12:39
mardydednick: mine doesn't because I'm reusing exactly the same fd, and not generating a new one12:39
mardydednick: so, I'll try calling mir_prompt_session_new_fds_for_prompt_providers() as I add a new app to the session, then if all works I'll close the bug12:41
mardydednick: mmm.... no, I'm actually requesting a new fd per each client... I'll investigate what it is, then12:43
=== dandrader|afk is now known as dandrader
=== MacSlow|lunch is now known as MacSlow
=== bremnes_ is now known as bremnes
cwayne-afkjanimo: just occurred to me we should probably discuss here instead :)13:31
=== cwayne-afk is now known as cwayne
janimocwayne, true :)13:35
cwaynejanimo: so im building a mako image now13:35
cwayneshouldn't take too long13:36
cwaynegot a pretty beefy desktop here :P13:36
ogra_cwayne, no, he is vega13:36
cwayneogra_: psh yeah right13:37
ogra_oh, i read "there"13:37
cwaynejanimo: done13:46
janimocwayne, built?13:46
cwayneseein stuff in out/target/product/mako13:48
janimocwayne, good.13:49
janimocwayne, you can now fastboot flash boot out/target/product/mako/boot.img13:49
janimoand fastboot flash recovery out/target/product/mako/recovery.img13:49
janimoor actually u-d-f could do that too if there is a working fastboot mode already13:50
cwaynejanimo: do i need to do the rootstock bits for the rootfs now?13:51
janimocwayne, I have never used that. I was wondering how to only do it via u-d-f. Need to make a device tarball first13:52
janimocwayne, a sec13:52
janimocwayne, until I put it in a proper repo:  http://paste.ubuntu.com/9214939/13:54
janimorun it by passing it a conf file as the first argument13:54
janimothat file should contain DEVICE=mako13:55
janimoyou can make the tar.xz by hand of course, making sure you have /system/var/lib/lxc/android/system.img in it13:56
janimobut this script is preferred, it's what we have been using internally for making device tarballs13:56
sergiusensjanimo: can't that just be a make target?13:56
janimosergiusens, I preferred not adding new targets to the build system, so we carry fewer patches (slightly different for each OEM tree) and to have easier dev and maintenance of this scipt13:57
janimoit's a separate step, but we can have one command builds by having a top level shell script do make + mktarball+ whatever other postprocessing13:58
cwaynecwayne@boomer:~/Projects/ports/opo/out/target/product/mako$ ubuntu-device-flash --device-tarball device_mako.tar.xz --bootstrap --channel ubuntu-touch/devel-proposed13:58
janimoit is what we do for some OEM builds ATM13:58
cwaynejanimo: ^ that look right?13:58
janimocwayne, you may need --device mako too13:59
janimo I think you need to be explicit when in bootstrap or recovery mode13:59
cwayneit got it somehow13:59
cwaynejanimo: stuck at google screen :/14:05
dednickmardy: yeah, i looked at the fd thing and it seemed ok on first glance. Only difference is that i think my map is calling "mir_prompt_session_new_fds_for_prompt_providers(X)" where x is the number of prompts to be added, rather than calling "X * mir_prompt_session_new_fds_for_prompt_providers(1)".14:07
Saviqmardy, hey, can you tell me if bug #1352251 is a problem still?14:08
ubot5bug 1352251 in unity8 (Ubuntu RTM) "Splash screen is shown as soon as QGuiApplication is instantiated" [High,Confirmed] https://launchpad.net/bugs/135225114:08
mardySaviq: yes, at least it was last time I checked14:09
mardySaviq: now I'm reflashing my device so I cannot check, but I'll do that tomorrow14:09
mardySaviq: is your question about whether the bug has already been fixed, or about whther it should get fixed?14:10
Saviqmardy, the latter14:11
Saviqmardy, as with trusted prompts you only create the Mir connection when you create the UI14:11
janimocwayne, it may be you need vendor blobs which we do not have in the repo14:12
Saviqsince you need to talk to the trusted socket (IIRC)14:12
janimoand last I checked it was not clear where to get tem in the wiki14:12
Saviqmardy, so I'm wondering whether this problem is still valid for you14:12
janimoI know I added that info last year14:12
janimocwayne, something like this, https://github.com/janimo/vendor_lge_mako but we need the 4.4 version now14:16
cwaynejanimo: like this? https://developers.google.com/android/nexus/drivers#mako14:16
janimocwayne, yes, those, not sure where those get unpacked, should be under /vendor/lge/mako14:17
cwaynewere on 4.4.2 right14:17
janimobut apart from the actual blobs google provides, some makefiles are needed too, to hook them in14:17
cwayneah, boo14:18
Saviqmardy, https://bugs.launchpad.net/qtmir/+bug/1352251/comments/714:19
ubot5Launchpad bug 1352251 in unity8 (Ubuntu RTM) "Splash screen is shown as soon as QGuiApplication is instantiated" [High,Confirmed]14:19
janimocwayne, this is the most annoying part of the build I agree14:19
janimocwayne, you could try building the kernel source for your device14:20
janimothe mako build was mostly for testing, it will probaly work once the vendor blobs are there14:21
cwaynejanimo: yeah, so maybe ill just skip mako14:21
cwaynejanimo: these two seem relevant to my interests: https://github.com/ubuntu-touch-oneplus-one14:22
cwaynejanimo: but i still don't understand like where to put these14:23
cwayneand how to get a lunch combo out of it14:23
cwayneand stuff like that14:23
janimocwayne, right, those along with the kernel should be the only device specific parts14:25
janimoone goes under device/oneplus/bacon14:26
cwayneso just git clone it there?14:26
cwaynethat'd be the proprietary_vendor_oneplus right14:26
janimothe other under vendor/oneplus/bacon14:26
janimocwayne, but the first step should be bringup of the Ubuntu kernel and recovery. This does not need the vendor bits14:27
janimothose are Android drivers and daemons for GPU and sensors and BT/Wifi which are only needed by the full system14:27
janimothe lunch combo is added by sourcing the .mk files under devices/ recursively14:28
cwaynealright, how do I do that?14:28
janimoand this repo should add that too14:29
janimocwayne, https://github.com/ubuntu-touch-oneplus-one/android_device_oneplus_bacon/blob/cm-11.0/vendorsetup.sh14:29
cwayneah, cool, except that it's for cm i guess14:29
janimocwayne, but this repo is for a CM build, which has a different build system than AOSP hence some files are not going to work14:29
* cwayne is worried this is all going to be too complicated for him14:30
janimocwayne, so you will need to adapt the .mk files but not other to look like those under device/lge/mako or others in AOSP14:30
janimocwayne, not complicated but tedious for sure14:30
janimoand error prone14:30
cwayneah, i can handle tedious probably :)14:30
cwayneok, so git clone the android_device_oneplus_bacon to device/oneplus/bacon?14:31
janimocwayne, yes14:31
cwaynejanimo: okies, done14:32
janimocwayne, but yes, it is a bit annoying to figure out which CM files to put in which AOSP file14:32
janimocm.mk is not needed at all for instance14:32
cwaynethen git clone propritery_vendor_oneplus_bacon to vendor/oneplus/bacon14:32
janimobut its content is14:32
cwayneah, okay14:33
cwayneso where do i move the content to14:34
janimocwayne, those files include each other, and only one (device_full.sh ?) is sourced by the build14:34
janimoBoarConfig.mk is needed14:34
janimocwayne, TBH I do not remember which does what, so I'd first have to read the mako ones to refresh my memory14:34
janimocwayne, you could try that, by trying to put similar definitions where they are in mako14:35
janimoor wait till I can have a look at this too later today14:35
janimocwayne, you could also try building the kernel which is orthogonal to all this14:35
cwaynei'll wait til you can have a look i suppose, I've got plenty of actual work to do :P14:36
janimoit should build as is (stock android), then add Ubuntu configs to it, then some Ubuntu patches (optional at this stage)14:36
janimocwayne, ok. Porting to new can  take weeks even for people who have done it before, so do not get discouraged14:36
janimohopefully we'll make that shorter by making the porting guide better and offering some tools to automate14:37
cwaynei'm happy to be the guinea pig for the new guide :)14:37
cwayneill be even happier if we get ubuntu on this beast of a phone14:37
janimocwayne, but this is sort of actual work too, if it makes our process more fluid and gets more porters on board :)14:37
cwaynejanimo: fair point :)14:37
janimojust not actual work assigned to you I guess :)14:37
janimoso I'll just ping you when I have better data and not have you waste too much time on this14:38
cwaynecool beans14:38
* cwayne kicks off a cm build anyway just to see if it works14:38
=== daker__ is now known as daker
denemehi i have an question about ubuntu tocuh15:22
denemecan in install it to mto g15:22
deneme*moto g15:22
ubot5You can find the full list of devices, official images, community images, and works in progress at https://wiki.ubuntu.com/Touch/Devices15:22
denemefor moto g it is working progress15:23
denemeso i cant install right now15:23
denemeam i right15:23
denemecan i try beta version something like that15:24
lotuspsychjedeneme: you can also try the XDA forums, maybe someone got it working15:24
popeyi love that quit message. makes me smile every time15:25
denemeit seems so risky15:25
lotuspsychjedeneme: well for now most supported devices are nexus15:26
lotuspsychjedeneme: perhaps the future will bring more support to devices15:26
denemewhat do u think about firefox os15:26
lotuspsychjenever tested it sorry15:26
denemeok ty15:27
=== chihchun_afk is now known as chihchun
elopiodiwic: ping. How can I know if a headset is plugged into the phone?15:59
sil2100Mirv: appmenu-qt5 should be good in the branch now16:01
sil2100(properly tested it with a qt 5.4.0-enabled PPA)16:01
diwicelopio, pactl list cards | grep available16:01
sil2100Mirv: I mean, it *builds* at least ;p16:01
sil2100Mirv: can't test it as I don't have a VM or a second machine to test it with 5.4.0 installed16:02
sil2100Mirv: but the code path should stay the same16:02
diwicelopio, or grep for the port name would probably be better, "grep -i headphone"16:03
diwicelopio, hmm, btw, not sure pulseaudio-utils is installed by default on the phone though...16:04
elopiodiwic: cool, that's useful. I would like to fake a headset, so I can test some things from the sound indicator.16:04
diwicelopio, in which case you'll have to use the native API instead16:04
elopiodiwic: I think I should do it using a dbus mock. Do you know of a better way to do it?16:05
=== dandrader is now known as dandrader|lunch
diwicelopio, I don't know what a "dbus mock" is, but I don'PulseAudio16:05
diwicelopio, I don't think PulseAudio exposes that informaiton over dbus16:05
diwicelopio, at least not on the phone16:06
diwicelopio, are you going to write a "fake PulseAudio" to test the sound indicator?16:06
elopiohum, that will make it harder.16:06
elopiodiwic: I don't want a fake pulseaudio if I can avoid it.16:07
elopiodiwic: I just want an easy way to make the phone think it has a headset plugged.16:07
diwicelopio, hmm16:07
diwicelopio, can you fake something in sysfs?16:09
diwicelopio, i e manually modify /sys/class/switch/h2w/state16:10
elopiodiwic: I cana try.16:10
diwicelopio, that's where PulseAudio reads it from on the phone16:10
=== chihchun is now known as chihchun_afk
diwicelopio, oh, but then you have to send uevents too...blah16:10
elopiodiwic: I see 2 with headphones, 0 without.16:11
diwicelopio, yup, that sounds right16:13
elopiodiwic: and I get an i/o error trying to write that file. And I would need root to make the change anyway, so that's not an easy way.16:13
elopioneeding root in the middle of the tests makes it complicated.16:13
diwicelopio, ok. So either you want to intercept between pulseaudio and the kernel, in which case we're talking about sysfs like we do now16:14
diwicelopio, or just hack pulseaudio. That's probably the simplest. the code is in src/modules/alsa/alsa-extcon.c16:15
mandelbarry, I pushed a number of changes for bug 1390205 can you take a look once the builds are done?16:41
ubot5bug 1390205 in ubuntu-download-manager "udm falls over on the switch from wifi to 3g meaning ota are stuck and the system can't recover till reboot" [Critical,In progress] https://launchpad.net/bugs/139020516:41
barrymandel: yep,thanks16:42
mandelbarry, no problem, took a little to find way it was happening16:42
seb128kenvandine, hey16:45
kenvandinehey seb12816:45
seb128kenvandine, did you notice that the flight mode/rotation lock/about settings listitems lost their icons in v?16:45
kenvandineseb128, no...16:46
seb128kenvandine, can you look if that happens to you?16:46
seb128just saw that on my desktop16:46
kenvandineyes, i just did16:46
kenvandinei see the same16:46
kenvandinei wonder what happened16:46
seb128I guess it's a toolkit regression16:46
seb128let me chase that down16:46
seb128I just wanted to check if you looked at the issue before starting16:47
kenvandineseb128, there was a ubuntu-ui-toolkit-theme update in image 2916:47
kenvandineseb128, yes... there was changes to Switch16:48
kenvandinewell SwitchStyle16:49
kenvandine"Use icons from theme"16:49
kenvandinebut this is really a ListItem right?16:49
=== chihchun_afk is now known as chihchun
seb128kenvandine, correct16:51
kenvandinewe use the icon from the suru theme16:52
kenvandinei wonder what other apps we can look at to compare16:52
kenvandineah, the uitk gallery :)16:52
kenvandineseb128, oh... it's all busted16:54
kenvandineany ListItem with an icon has no icon16:55
kenvandineeven in the uitk gallery16:55
kenvandinebzoltan, ^^^ is that known?16:55
=== dandrader|lunch is now known as dandrader
bzoltankenvandine: where is it?16:58
kenvandinebzoltan, seb128 just filed bug 139579316:59
ubot5bug 1395793 in ubuntu-ui-toolkit (Ubuntu) "Icons missing in ListItems widgets" [Undecided,New] https://launchpad.net/bugs/139579316:59
seb128bzoltan, it's a regression from friday's toolkit update, downgrading the deb fixes it17:00
kenvandineso has that version landed in rtm as well?17:01
seb128I doubt it17:02
bzoltankenvandine:  no, we do not land UITK to RTM for long time17:02
kenvandineok... i know vivid had been way behind rtm17:03
bzoltankenvandine:  to RTM we land only for special request17:04
bzoltanseb128: kenvandine: is there any app what uses icons in listitems?17:05
seb128bzoltan, system settings17:05
kenvandineand the uitk showcase :)17:05
kenvandineat least17:05
kenvandinei'm sure there are others17:05
seb128messages could be17:06
seb128contacts as well17:06
seb128I didn't look at the code but their UI is list like with text and icons on the left17:06
bzoltanseb128:  where the system setting is using it?17:09
seb128bzoltan, on the main screen for the orientation lock/flight mode/about this device/reset device items17:09
bzoltanseb128:  OK, I am flashing the device and check it17:10
seb128bzoltan, you can test the small testcase from the bug if you want17:10
seb128on desktop17:10
bzoltanseb128:  I am not on vivid, yet17:11
seb128bzoltan, also cf #sdk discussion, I'm testing a branch from kalikiana which might fix it17:11
seb128bzoltan, you don't run the current toolkit version on your utopic?17:11
bzoltanseb128:  of course not17:12
seb128bzoltan, the concept of not running what you work on seems weird to me, but ok ;-)17:13
bzoltanseb128: hmm...17:13
bzoltanseb128:  target env !=  dev env17:14
seb128bzoltan, well, for sure I've ways/chroots to test all series of the stuff I work on17:14
seb128or VMs17:14
bzoltanseb128:  we have emulators and devices as targets, we use click chroot as rootfs for development17:14
bzoltanseb128:  I have both emulators and devices on vivid ...17:14
seb128bzoltan, yeah, and I'm surprised you don't have an handy way to run a test program against the trunk version of the toolkit17:14
seb128like you need to actually reinstall a device to test trunk17:15
bzoltanseb128:  but it is a very broadly spread misconception that the development environment should be runtime environment too17:15
seb128I though you would have an emulator ready to boot17:15
bzoltanseb128:  I have super handy way .. emulator17:15
seb128bzoltan, I never said that17:15
seb128so why don't you boot the emulator and try the 10 line qml example from the bug?17:16
bzoltanseb128:  there is no update mechanism for emulators... so I need to create an emulator what _might_ have the image with the UITK released to Vivid on Friday17:17
bzoltanseb128:  I had the latest UITK with Vivid on my device, but I have flashed it with RTM image today for other tests ... my other device is on Vivid I flashed last week.17:18
seb128bzoltan, ok, let's stop that discuss here, I find it fascinating that you don't have a trivial way to run a test program against trunk of the uitk but it's not my issue17:18
bzoltanseb128:  I do have...17:19
seb128it feels like that's the sort of things you should have wrapper/tools that let you do that in 10 seconds at any time17:19
kalikianaseb128: testing a branch of mine?17:19
seb128kalikiana, https://code.launchpad.net/~ubuntu-sdk-team/ubuntu-ui-toolkit/imageSourceNotNOTIFYable/+merge/24265517:19
seb128kalikiana, it seems to fix https://bugs.launchpad.net/ubuntu/+source/ubuntu-ui-toolkit/+bug/139579317:19
ubot5Launchpad bug 1395793 in ubuntu-ui-toolkit (Ubuntu) "Icons missing in ListItems widgets" [High,New]17:19
kalikianaseb128: hmmm if that's a side effect, I'd like to see if we can unit-test that explicitly17:20
bzoltanseb128:  I have like 6 different vivid/rtm setups ... I am testing branches and the staging like 24/7 ... in a bug report like this the first thing i do is a clean flash and brand new emulator. To be sure that I am testing the right stack.17:20
seb128bzoltan, k, I think your workflow is not efficent, you should maybe have a vivid installation in a VM, so you would just have to boot the VM and qmlscene the testcase, that would give you a trivial 15 seconds work way to confirm a bug17:21
bzoltanseb128:  When I have seen your report, I started to create a new emulator and flashing a device ... that is how fast I can do it.  My desktop is Utopic, because the target of the SDK tools _IS_ Utopic and Trusty.17:21
bzoltanseb128:  I have limited number of computers and limited resources .. I am testing the tools on Trusty and Utopic... testing the RTM and Vivid on devices/emulators.17:22
bzoltanseb128:  I am happy to explain to you how I work and why I choose certain workflow. I am happy to hear your suggestions... it is for sure possible to improve.17:23
seb128bzoltan, well my suggestion is easy, have a desktop VM with vivid you can boot easily17:23
seb128it's sometime easier to debug/test on a desktop than on a device17:24
bzoltanseb128: Simple said... The UITK I am testing on emulator and device, because there are the targets, not the desktop. The desktops I am using for tools validation... that is why I am using U and T on desktop.17:24
seb128but then it's just a suggestion, feel free to ignore it17:24
seb128bzoltan, well, most issues can be reproduced and debugged on desktop17:24
seb128bzoltan, like the one I reported, I reported it from a desktop and already confirmed that kalikiana's vcs fixes it17:24
seb128I didn't test on a device but I'm pretty confident that if the fix works on desktop it's going to work on the device17:25
bzoltanseb128:  I do not confirm bugs based on desktop only tests. So an emulator and device  test is mandatory for me.17:25
mandelbarry, packages ready => v17:49
mandelbarry, https://code.launchpad.net/~mandel/ubuntu-download-manager/adapt-network-changes/+merge/24208317:49
barrymandel: k, thx, might be a little while17:52
mandelbarry, I have tested with the current vivid proposed and looks ok (gets downloaded and installed)18:05
mandelbarry, yet the UI is utterly broken, I need to fix that crap18:05
=== alan_g is now known as alan_g|EOD
sil2100Mirv: if you could test this branch and approve/top-approve it, would be great ;)18:17
=== chihchun is now known as chihchun_afk
plarsogra_: so I guess due to https://bugs.launchpad.net/ubuntu/+source/android-tools/+bug/1382559 we really need the new udf now? (recall we previously reverted it to the previous version)19:12
ubot5Launchpad bug 1382559 in android-tools (Ubuntu RTM) "adbd does not check the screen state before allowing a connection" [Critical,Confirmed]19:12
ogra_plars, oh, yeah19:12
ogra_but that also means rtm needs the new recovery19:13
ogra_or you keep the old UDF for rtm19:13
ogra_since the change will only land in OTA-119:13
plarsogra_: yes, so we could still get to a bad state on our devices with rtm if use the new udf, and we will get them in an unreachable state if we use the old udf with vivid19:13
plarsogra_: oh, so it's not in the vivid images yet?19:14
ogra_i suspect you wont to keep them distinct somehow anyway19:14
ogra_they will be diverted all the time19:14
plarsogra_: what do you mean by distinct?19:14
plarsogra_: we don't use a different udf for different images19:14
ogra_plars, well, i suspect you will have to in the future19:15
ogra_devel and rtm will diverge more and more19:15
ogra_especially in recovery features i fear19:16
plarsogra_: that doesn't seem reasonable? do we really expect to need a different udf for them for any other reason than this bug? UDF isn't very friendly to run in this way19:16
ogra_well, many fixes and changes are held back from RTM19:16
ogra_and that amount will raise19:16
plarsogra_: I understand rtm is locked down, but couldn't this be handled in udf in some way that's compatible with either? or if not, then vivid could be made to match what's in rtm?19:17
ogra_probably ... sergiusens ?19:17
plarsotherwise, we always run the risk of bricking a lot of devices19:17
ogra_i really suspect that this will only be the first time where you have some heavy divergence though19:18
ogra_rtm is a different distro ...19:18
plarsogra_: true, but I really don't think you guys want to maintain different tools for flashing different channels any more than the users of those tools (including ci) want to try to sort out which version has to be used with which channel19:21
plarsso one version of the tool really has to work with all channels, with good deprecation if an api change is needed19:21
ogra_i think sergiusens had something like that in the drawer19:22
ogra_(deprecation test plans etc ... didnt we have a meeting at teh sprint ?)19:22
sergiusensogra_: it is implemented19:22
sergiusensogra_: but, as I said, the recovery pushed to rtm is broken19:23
sergiusensogra_: so it needs updating19:23
sergiusensbefore anything can be done19:23
ogra_that wont happen soon though19:23
sergiusensplars: the tool is deprecated correctly; this is the same problem we talk about every week19:24
sergiusensogra_: it's really a one line fix that needs to land into rtm19:24
ogra_sergiusens, yes, that will not happen for at least another week19:25
sergiusensogra_: well then this needs to wait until then19:25
* ogra_ double sighs ... 19:25
sergiusensogra_: not much we can do, really19:25
ogra_well, then let me roll back mandel's changes19:26
ogra_but later tonight, i wanted to stop working hours ago19:26
plarsogra_: sorry to bring bad news on that, just trying to avoid a fire drill when the next image lands :(19:26
ogra_plars, not your fault really ... i just ahd a really busy and rather bad day19:27
sergiusensogra_: bad way to start the week indeed19:35
ogra_sergiusens, yeah, i should rather move with the fashion and develop stones in some organ instead :)19:41
sergiusensogra_: believe me, that brings in a crappier situation than just a bad day ;-)19:43
ogra_thats what i meant ... in that light a bad day still seems ok19:44
elopioping charles: I flashed 15.04 in my krillin, image #37. And when I play music, the sound indicator doesn't show the play/pause controls.20:07
elopiois that supposed to work?20:07
charleselopio, yeah it should. Could you open a ticket for it and assign it to me?20:08
elopiocharles: sure.20:08
charleselopio, are you doing anything unusual to play the music?20:08
elopiocharles: no. I have one song in the music folder. I open the music player, and play it.20:09
charlesok, sounds like a straightforward use case :)20:11
elopiocharles: https://bugs.launchpad.net/ubuntu/+source/indicator-sound/+bug/139586320:13
ubot5Launchpad bug 1395863 in indicator-sound (Ubuntu) "player controls are not displayed in the sound indicator while music is playing" [Undecided,New]20:13
charleselopio, thanks20:13
elopiocharles: the indicator detects if a song is playing calling mpris, right?20:13
charleselopio, right20:14
ahayzenelopio, was this reenabled in vivid? as i know it was disabled/reverted in rtm ?20:14
elopioahayzen: I don't know.20:14
charleselopio, though tbh, ted's dug in this code more recently than me; he may be able to give more educated answers than me20:14
charleselopio, which is partially the reason I wanted the bug; it's been too long since I've dug into that code20:15
ahayzenelopio, because there were multiple bugs/features missing so it was reverted for rtm20:15
elopioahayzen: I see. Do you have a bug number for that?20:15
ahayzenelopio, for the revert? or just the issues lol20:15
charleselopio, oop, we're on freenode, s/ted/tedg/20:16
elopioahayzen: for the revert.20:17
ahayzenelopio, i'm searching and can't find it... in bug 1378048 Pat mentions that they are hidden20:18
ubot5bug 1378048 in media-hub (Ubuntu) "Make music controls work in the sound indicator" [Critical,Confirmed] https://launchpad.net/bugs/137804820:18
ahayzenelopio, i'm not sure if there was a bug for the revert in the end...i think it just happened due to the number of bugs/features missing20:18
elopiook, thanks. That's useful info.20:20
tedgelopio, It uses MPRIS, but the media service no longer exports that.20:27
tedgelopio, You can set an environment variable to turn it on, but the implementation is incomplete.20:27
elopiotedg: what env var is that?20:28
tedgelopio, I don't remember, if you look at the media hub changelog I'm sure it's there.20:29
tedgkenvandine, So it seems there's still one test failing, but I'm not sure why. Any ideas? https://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-runner-vivid-mako/304/?20:36
tedgkenvandine, Is it getting more than one system settings instance?20:36
barrymandel: if you're still around.  tested and verified.  commented on the mp.  thanks!20:38
kenvandinetedg, i doubt it, i'll look20:42
kenvandinetedg, i think this is an example of why we need to refactor lots of tests20:43
tedgHeh, I just want to land my little patch, I don't care about refactoring tests! :-)20:44
kenvandinetedg, does this test pass on your device?20:45
kenvandinetedg, and have you had this error consistently?20:45
tedgkenvandine, I haven't done that with teh session change, let me try.20:45
kenvandinei suspect it might be environmental20:45
kenvandineand maybe just flaky20:45
kenvandinepull: /var/crash/_usr_bin_autopilot3.32011.crash -> /var/lib/jenkins/slaves/mako-05/workspace/generic-deb-autopilot-runner-vivid-mako/clientlogs/ubuntu_system_settings/_usr_bin_autopilot3.32011.crash20:47
kenvandinetedg, whatever happened, it caused autopilot3 to crash20:48
veeberskenvandine: Hmm, just had a quick look now "return codecs.latin_1_decode(input,self.errors)[0] -> TypeError: 'NoneType' does not support the buffer interface"20:50
veebersthomi: You were working on adding something to test tools to make debugging this type of error easier, right? ^^20:51
kenvandinetedg, unity8 crashed20:51
kenvandinethat was probably the root cause20:51
tedgAh, hmm.20:51
tedgI'm pretty sure that wasn't my test's fault :-)20:51
* thomi looks20:51
kenvandineno ubuntu-system-settings crash file though20:52
thomiveebers: wait, what am I looking at? https://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-runner-vivid-mako/304/testReport/junit/ubuntu_system_settings.tests.test_sound/SoundTestCase/test_sound_page/ seems pretty straight-forward to me?20:52
kenvandineat least for me, that kind of error i need to back my way through the rest of the artifacts20:53
kenvandinethat wasn't obvious to me that unity8 crashed20:53
veebersthomi: hmm perhaps I jumped the gun. I was being nosey at the autopilot crash file kenvandine suggested (Looking at https://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-runner-vivid-mako/304/artifact/clientlogs/ubuntu_system_settings/_usr_bin_autopilot3.32011.crash)20:53
tedgOh, uhg. RTM and those packages are vivid.20:53
kenvandineyeah, this CI run is vivid20:54
thomiveebers: right, basically someone's tried to add None as a text content, which is obviously not correct.20:55
thomiveebers: I have an open work item to make testtools raise an error when you add the content item, but I've not had time for that20:56
veebersthomi: ah right, that's what I had in mind. sweet20:56
=== salem_ is now known as _salem
mandelbarry, looking21:17
tedgUhg, nothing passes on my device.21:27
tedgkenvandine, They all fail on my device :-/21:31
tedgkenvandine, Not sure what to do with that result.21:31
kenvandinetedg, pastebin please21:32
tedgkenvandine, http://paste.ubuntu.com/9220998/21:34
kenvandineinstall that21:34
kenvandinethomi, shouldn't something have a depends on that?21:35
tedgkenvandine, Already have the newest version.21:35
thomikenvandine: autopilot-touch should depend on it, IIRC21:35
kenvandine21:23:36.549 WARNING testcase:538 - Failed to create Touch device for bug lp:1297595 workaround: No module named 'evdev'21:36
thomitedg: what does 'python3 -m autopilot.run -v' show?21:37
tedgthomi, phablet@ubuntu-phablet:~$ python3 -m autopilot.run -v21:38
tedgAutopilot Source Version: 1.5.0 Autopilot Package Version:21:38
kenvandinetedg, and you're sure your device has python3-evdev right?21:38
kenvandinenot looking for python-evdev21:38
kenvandinealthough that shouldn't be there21:39
thomithat looks correct to me. kenvandine is suggesting what I'd check next21:39
kenvandineUInput: ImportError("No module named 'evdev'",)21:39
tedgUhg, you're right, I didn't put a 3 there.21:39
* tedg hates python 2.721:39
tedgWhen can we remove it from the archive?21:39
kenvandinetedg, really? i'm surprised you added a version21:40
kenvandineyou're such a hater :)21:40
thomiI was going to say the same thing!21:40
thomitedg: you've got a reputation now :D21:40
* tedg puts the snakes on a plane with Samuel L. Jackson.21:41
=== greyback__ is now known as greyback
tedgkenvandine, Okay, so all the tests passed.21:48
tedgNot sure what the means for Jenkins though.21:49
kenvandinekick a rebuild in jenkins please21:49
=== _salem is now known as salem_
SturmFlutmzanetti, popey: I rolled out mzanetti's last commit, the app store feed has pictures now. Works at least in Thunderbird and gReader22:26
SturmFlutYou may have to refresh your caches22:27
SturmFlutdepending on the application22:27
* ogra_ tries in liefrea22:29
ogra_nice !!22:30
SturmFlut> grep appstorediff /var/log/apache2/access.log | wc -l22:32
SturmFlut> 77122:32
mzanettiSturmFlut: hmm... isn't that just the amount of refreshes?22:33
SturmFlut771 requests since yesterday. So this is what internet stardom feels like22:33
mzanettihmm... wonder how often thunderbird refreshes22:34
SturmFlutmzanetti: I see a lot of different IPs, and surprisingly many over IPv622:34
mzanettithat's me22:34
SturmFlutand myself ;)22:34
* mzanetti has a lot of troubles atm with his IPv6 connection22:35
mzanettialthough its probably that shitty modem/router22:35
SturmFlut> grep appstorediff /var/log/apache2/access.log | awk '{print $1}' | sort | uniq | wc -l22:35
SturmFlut> 10722:35
SturmFlutmzanetti: Okay, found the "culprit"22:36
mzanettiok. now filter out IPv6 address randomization and the daily switching IPv4 addresses of the others :D22:36
SturmFlutmzanetti: Do you know a guy named "Google WebCrawler"? He seems to like Ubuntu Touch a lot22:37
popeySturmFlut: you'll see at least 3 IPs for me, home, vps and a hotel in istanbul ☻22:37
mzanettiso in the end it's probably just me, popey, the google crawler and yourself causing all the traffic22:38
ogra_popey, oh, *that* turkey ... i thought you fly to the US for thanksgiving parties when you said "off to turkey" :)22:38
SturmFlutso this is what it feels like to lose internet stardom22:39
ogra_its only two days yet, aint it ?22:39
mzanettiwell, it's up since a month or two, but only "announced" since two days, yeah22:39
SturmFlutogra_: I was expecting "Gangnam Style" levels of stardom22:39
ogra_oh, billions of hits then ... :)22:40
* SturmFlut was not joking about that monster server22:40
* SturmFlut never jokes22:40
* mzanetti confirms22:40
mzanettiwhat the!!!22:42
mzanettiphablet-screenshot only creates screenshots of the greeter, even though I see the device unlocked22:42
mzanettifail of the week22:43
mzanettiSturmFlut: works in shorts too: http://i.imgur.com/315FVhL.png22:44
popeyhaha ogra_22:45
popeySturmFlut: not seeing images in the shorts app here. forced a refresh but still no.22:46
SturmFlutmzanetti: I wish I had more time to work on it. But this week is again filled with business trips.22:46
popeyand now i do!22:46
mzanettiSturmFlut: no worries... all is fine22:46
popeyhow cool is that22:46
popeyreading the rss feed for the store, on the phone22:47
=== salem_ is now known as _salem
mzanettipopey: what version of shorts is that?22:47
mzanettilooks like this here: http://i.imgur.com/315FVhL.png22:47
popeysecret version :D22:47
popeyoh change your view22:47
popeyyou're in list view, change to grid view22:48
popeythe only difference I have is colour22:48
mzanettiyeah, grid works too22:48
mzanettipopey: and UbuntuShape--22:48
popeyfeel free to review the branch :D https://code.launchpad.net/~qqworini/ubuntu-rssreader-app/color-experiment22:49
SturmFlutmzanetti: I do get bored by all those supercomputers sometimes ;)22:50
mzanettido you have access to them from home? like in a way you can use them as private playground?22:52
SturmFlutmzanetti: Time on production machines is scarce, so it would be unfair to steal resources from scientific users. But we do have a lot of test clusters, and I use them a lot to test various things, also from home.22:54
mandelrobru, the tests in silo 15 have to be done in a later img to ensure that we can upgrade :)22:54
robrumandel: are you saying I shouldn't publish it then?22:55
SturmFlutFor example I currently use a machine with 40 cores and 512 GB of RAM as my build machine22:55
mzanettithat'll do23:00
SturmFlutmzanetti: I'll visit http://www.fz-juelich.de/ias/jsc/EN/Expertise/Supercomputers/JUQUEEN/Configuration/Configuration_node.html this week. There are only seven systems in the world which are faster.23:01
sarnold"Operating system: RedHat Linux V6.2"23:03
sarnoldis that really RHL or is that actually RHEL? heh23:03
SturmFlutsarnold: RHEL, sadly23:04
* mzanetti would be happy with an Orange Box :)23:04
sarnoldmzanetti: mmmm :)23:04
sarnoldSturmFlut: at least RHEL is from this decade :)23:04
SturmFlutmzanetti: Actually the Orange Box can easily be outperformend by a single dual-socket Xeon system, and the single system is even cheaper23:05
mzanettiyeah... mostly for learning purposes... I don't really have that many production use cases for this stuff :)23:06
mzanettimy laptop seems fast enough to compile unity8, which is what I do 90% of the time anyways :)23:06
SturmFlutsarnold: Supercomputers tend to run very recent operating systems. You need support for bleeding-edge hardware and the product cycles are very short. RedHat and SuSe have to backport an incredible amount of drivers into their enterprise kernels to make sure that they still run on all the newest enterprise hardware.23:08
* mzanetti would think one doesn't buy a new supercomputer every year :D23:08
SturmFlutmzanetti: ...actually, we do23:08
sarnoldSturmFlut: well, those were 'interaction nodes', I wondered if they juts got something working a decade ago and stuck with it even as they had to upgrade the software for the new hardware for the compute and io nodes along the way23:08
SturmFlutmzanetti: ...actually, we didn't buy one supercomputer this year, we bought two23:09
SturmFlutI do feel like a crazy person sometimes when I write/say stuff like this23:10
SturmFlutsarnold: It depends. Crazy people like Fujitsu, IBM, SGI and Cray build crazy systems. They run different operating systems on the login nodes than on the compute nodes, they have dedicated IO nodes, they have strange networks etc. Averybody else just buys a huge number of nodes with the same CPUs, installs RHEL or SLES and clusters everything together.23:13
sarnoldSturmFlut: I did wonder if that was relatively 'commodity' or not; the 40gbps 5d torus isn't much vote one way or another but the 2.5 μsec latency sounds amazing... is that something you can get with standard ethernet sff connectors or does it require something much more fun? :)23:15
mandelrobru, no, I'm saying you should published, just that if you look at the number of the img used is longer on purpose :)23:20
mandelrobru, is not that I did not test it in the latest due to an error from my side23:20
robrumandel: oh ok. I thought you were saying we had to wait for the next image to come out in order to test it fully.23:21
robrumandel: ok I'll publish it23:21
SturmFlutsarnold: You can get down to 1,6 µsec with 40G Ethernet, and less than one µsec with FDR (56G) InfiniBand, but when you have to adress 50.000 nodes, your network gets so big that the number of switches between two nodes starts to add a lot of latency. 2,5 µsec worst-case latency is pretty good.23:21
=== johnlage is now known as johnlage_partyha
SturmFlutsarnold: We use InfiniBand without exception, Ethernet is just crappy technology IMO23:22
mandelrobru, the other way, that you can publish without fear that I tested it in the wrong img by accident, was done on purpose23:23
mandelrobru, sorry if I repeat myself, bip is misbehaving23:23
robrumandel: alright, thanks for clarifying ;-)23:23
sarnoldSturmFlut: do you do IP over IB? or do you use something else?23:55

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