--- Log opened Fri Aug 23 00:00:27 2013 | ||
-!- travis-ci [~travis-ci@ec2-54-224-203-225.compute-1.amazonaws.com] has joined #shogun | 00:06 | |
travis-ci | [travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/10513097 | 00:06 |
---|---|---|
-!- travis-ci [~travis-ci@ec2-54-224-203-225.compute-1.amazonaws.com] has left #shogun [] | 00:06 | |
-!- sonne_ [~sonne@f053035226.adsl.alicedsl.de] has joined #shogun | 00:11 | |
sonne_ | wiking: we had modules with swig | 00:13 |
-!- sonne_ is now known as sonne|osx | 00:13 | |
sonne|osx | wiking: but this caused all kinds of trouble (crashers/stuff not working) due to dependencies between modules | 00:13 |
sonne|osx | wiking: that is why I dropped that maybe about 1-2 years ago | 00:14 |
shogun-buildbot | build #1466 of deb2 - static_interfaces is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/1466 blamelist: Soeren Sonnenburg <sonne@debian.org> | 00:14 |
sonne|osx | ohh and compile time was actually not that much shorter... modules are bigger due to quite some overlap (e.g. shogun.Kernels needed shogun.Features) | 00:15 |
sonne|osx | and yes that is why we have these shogun.Features etc - we started modular | 00:15 |
shogun-buildbot | build #31 of osx1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/osx1%20-%20libshogun/builds/31 blamelist: Soeren Sonnenburg <sonne@debian.org> | 00:18 |
shogun-buildbot | build #1497 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1497 blamelist: Soeren Sonnenburg <sonne@debian.org> | 00:19 |
shogun-buildbot | build #963 of rpm1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/963 | 00:26 |
shogun-buildbot | build #1287 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1287 blamelist: Soeren Sonnenburg <sonne@debian.org> | 00:42 |
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 245 seconds] | 00:42 | |
-!- pickle27 [~kevin@d67-193-243-174.home3.cgocable.net] has left #shogun [] | 02:03 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 02:21 | |
shogun-buildbot | build #1581 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1581 blamelist: Soeren Sonnenburg <sonne@debian.org> | 02:31 |
-!- sonne_ [~sonne@f053035142.adsl.alicedsl.de] has joined #shogun | 02:59 | |
-!- sonne|osx [~sonne@f053035226.adsl.alicedsl.de] has quit [Ping timeout: 245 seconds] | 03:01 | |
-!- sonne__ [~sonne@f053046196.adsl.alicedsl.de] has joined #shogun | 03:59 | |
-!- sonne_ [~sonne@f053035142.adsl.alicedsl.de] has quit [Ping timeout: 276 seconds] | 04:02 | |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has quit [Ping timeout: 241 seconds] | 06:30 | |
-!- sonne__ [~sonne@f053046196.adsl.alicedsl.de] has quit [Quit: sonne__] | 07:16 | |
-!- zxtx [~zv@cpe-76-166-29-100.socal.res.rr.com] has joined #shogun | 07:31 | |
-!- gsomix_ is now known as gsomix | 07:40 | |
gsomix | chess game on sed. VERY NICE https://github.com/bolknote/SedChess/blob/master/chess.sed | 07:40 |
-!- sonne_ [~sonne@89.204.137.122] has joined #shogun | 08:08 | |
-!- sonne_ [~sonne@89.204.137.122] has quit [Quit: sonne_] | 08:28 | |
-!- sonne|work [~sonnenbu@91-64-72-127-dynip.superkabel.de] has joined #shogun | 09:08 | |
sonne|work | wiking: I am trying to do a cmake summary page | 09:22 |
-!- zxtx [~zv@cpe-76-166-29-100.socal.res.rr.com] has quit [Ping timeout: 260 seconds] | 09:29 | |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun | 09:36 | |
-!- mode/#shogun [+o lisitsyn] by ChanServ | 09:36 | |
-!- gsomix_ [~gsomix@178.45.75.171] has joined #shogun | 09:57 | |
-!- gsomix [~gsomix@80.234.19.54] has quit [Ping timeout: 264 seconds] | 10:00 | |
@wiking | sonne|work: \o/ | 10:02 |
sonne|work | wiking: do you have an idea how I can match the name of a variable? | 10:04 |
sonne|work | wiking: I mean say I iterate over all variables | 10:04 |
sonne|work | but want to print out stuff only for variables with name starting with USE_ | 10:05 |
sonne|work | I guess I have it | 10:05 |
@lisitsyn | sonne|work: get_cmake_property(vs VARIABLES) | 10:08 |
@lisitsyn | then | 10:08 |
@lisitsyn | foreach(v ${vs}) | 10:09 |
@lisitsyn | ... check if USE_ and some actions | 10:09 |
@lisitsyn | endforeach(v) | 10:09 |
@lisitsyn | I am searching how to check it | 10:09 |
@lisitsyn | sonne|work: | 10:12 |
@lisitsyn | if ($v MATCHES "your regexp") | 10:12 |
@lisitsyn | ... your actions | 10:12 |
@lisitsyn | endif() | 10:12 |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun | 10:24 | |
-!- iglesiasg [~iglesias@2001:6b0:1:1041:5158:760a:da84:6c51] has joined #shogun | 10:24 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 10:24 | |
sonne|work | lisitsyn: yea found it already... | 10:24 |
@iglesiasg | good morning | 10:25 |
@lisitsyn | iglesiasg: hey | 10:26 |
@lisitsyn | sonne|work: where are you going to stay in the us? | 10:32 |
sonne|work | lisitsyn: errm in the hotel :D | 10:32 |
@lisitsyn | sonne|work: but you are arriving a week before? | 10:33 |
sonne|work | lisitsyn: no | 10:33 |
@lisitsyn | when will you arrive? | 10:33 |
sonne|work | lisitsyn: I wanted to arrive on friday just in time for dinner | 10:33 |
@lisitsyn | friday 18th? | 10:33 |
@lisitsyn | sonne|work: so something changed I guess? I was sure you said you planned to arrive earlier | 10:34 |
sonne|work | no nothing changed | 10:34 |
sonne|work | I said that all the time... | 10:34 |
@lisitsyn | ah may be you stay more? | 10:34 |
@lisitsyn | when you go back then? | 10:34 |
@lisitsyn | sonne|work: it can't be I play with my imagination :D | 10:35 |
sonne|work | My plan is/was to be there on the 18th and then stay one week | 10:35 |
@lisitsyn | ooh alright | 10:35 |
@lisitsyn | where do you stay next week then? | 10:35 |
sonne|work | lisitsyn: well I think you are pretty overloaded these days... | 10:35 |
@lisitsyn | sonne|work: I just wanted to clarify what do you and Heiko plan :) | 10:36 |
sonne|work | No idea | 10:37 |
sonne|work | I think on the day it became clear I sent an email to you 2 with my plans but never got a reply | 10:38 |
sonne|work | lisitsyn: will you stay longer than the summit? | 10:38 |
sonne|work | lisitsyn: or will you just arrive on the 18th and go back on the 20th? | 10:38 |
@lisitsyn | sonne|work: I haven't decided but it wouldn't make sense if I would be 'alone' :) | 10:39 |
sonne|work | alone as in w/o gf? | 10:39 |
@lisitsyn | sonne|work: w/o you and Heiko too | 10:39 |
@lisitsyn | :D | 10:39 |
@lisitsyn | I mean I don't know anybody there | 10:40 |
@iglesiasg | lisitsyn, perfect time to make friends in the US! | 10:41 |
@lisitsyn | iglesiasg: haha how? just walk around and ask people? | 10:41 |
@lisitsyn | ;) | 10:41 |
sonne|work | I guess we could meet Olivier and Quoc | 10:41 |
sonne|work | lisitsyn: you have to talk to all the people at the summit | 10:42 |
sonne|work | and done | 10:42 |
@lisitsyn | yes sure but still | 10:42 |
@iglesiasg | lisitsyn, the mentors gsocers | 10:42 |
@lisitsyn | sonne|work: I don't mind to stay a few days more | 10:42 |
sonne|work | they are all in the same boat | 10:43 |
@lisitsyn | sonne|work: so we can travel around and hack shogun or whatever | 10:44 |
@lisitsyn | that's why I ask ;) | 10:44 |
sonne|work | lisitsyn: well anything is possible... I will certainly rent a car | 10:44 |
sonne|work | so we could go hiking in the redwoods or some other national park | 10:45 |
sonne|work | etc | 10:45 |
@lisitsyn | sonne|work: I found the mail so I am recalling you to get a new passport | 10:45 |
sonne|work | yeah already done (will get one within a month latest) | 10:46 |
sonne|work | lisitsyn: and don't forget to apply for esta http://www.esta.us/ | 10:46 |
@lisitsyn | sonne|work: I didn't know about it | 10:47 |
@lisitsyn | sonne|work: no it is not for me | 10:47 |
* wiking is doing hadoop integration of shogun | 10:48 | |
sonne|work | I guess you need a passport for doing so | 10:48 |
@lisitsyn | sonne|work: it is only for you and heiko (visa waiver programme) | 10:48 |
@lisitsyn | I will have a passport with an us visa | 10:48 |
@lisitsyn | (if they not reject me) | 10:48 |
@iglesiasg | wiking, it sounds cool! What do you want to use it for? | 10:48 |
@wiking | iglesiasg: now? nothing... i just want to have it so that it can be serialised to and from hdfs + do some jobs over hadoop | 10:49 |
@lisitsyn | sonne|work: anyway thanks for letting me know | 10:51 |
@lisitsyn | I will have B2 visa | 10:52 |
@lisitsyn | (I hope) | 10:52 |
@lisitsyn | ah no B1 | 10:53 |
@wiking | iglesiasg: which reminds me to do automated test with serialization + the ones heiko asked for | 10:54 |
sonne|work | wiking: and upload the videos and do static interfaces w/ cmake :D | 10:55 |
sonne|work | wiking: did you know that you lead the stats this month? | 10:55 |
@wiking | sonne|work: hehe good on me \o/ | 10:56 |
@wiking | sonne|work: yeah about the statics | 10:56 |
@lisitsyn | sonne|work: wiking lead everything this month | 10:56 |
@wiking | sonne|work: you might know things the best | 10:56 |
@wiking | sonne|work: because i really hate digging in the shogun's old makefile.template | 10:56 |
sonne|work | lisitsyn: btw did you notice that you removed 3 times as much code as you added? | 10:57 |
sonne|work | err 10 times | 10:57 |
@wiking | sonne|work: python_static = libshogun objs + the python_static objs in a shared lib? | 10:57 |
sonne|work | wiking: exactly | 10:57 |
@wiking | sonne|work: r_static the same? | 10:57 |
@wiking | sonne|work: matlab_static? | 10:57 |
sonne|work | wiking: all the same | 10:58 |
@wiking | and while we are at it, i dont understand this at all: | 10:58 |
@wiking | http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1581/steps/compile%20csharp_modular/logs/stdio | 10:58 |
@wiking | sonne|work: ^ | 10:58 |
sonne|work | wiking: only elwms will need to link against all | 10:58 |
sonne|work | wiking: that is an issue with cshapr and typemaps | 10:58 |
@wiking | sonne|work: but why only on slave3? | 10:58 |
sonne|work | wiking: when you do a function foo(SGVector,SGVector) csharp will vomit | 10:58 |
@lisitsyn | sonne|work: yeah I have seen | 10:59 |
@lisitsyn | no idea why | 10:59 |
@wiking | sonne|work: slave4 works ok | 10:59 |
sonne|work | wiking: no way | 10:59 |
@lisitsyn | may be I AM THE DESTRUCTOR | 10:59 |
@wiking | sonne|work: yes way: http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1579/steps/test%20csharp_modular/logs/stdio | 10:59 |
sonne|work | lisitsyn: your contry is produced the best destructors the world has | 10:59 |
sonne|work | wiking: no way really | 10:59 |
sonne|work | wiking: must be different branch or sth | 10:59 |
@lisitsyn | sonne|work: yeah years ago | 11:00 |
@wiking | sonne|work: dunno why but slave4 compiles ok and slave3 does not | 11:00 |
sonne|work | wiking: or some define not active or include missing - cannot be | 11:00 |
@wiking | oh osx bot is down :( | 11:01 |
@wiking | sonne|work: do we still want cmd_static? :D | 11:01 |
sonne|work | wiking: yes | 11:03 |
@wiking | ... | 11:03 |
@lisitsyn | sonne|work: have you seen NIPS is hosting MLOSS workshop? | 11:04 |
sonne|work | lisitsyn: is there an official workshop list / announcement somewhere? | 11:05 |
@lisitsyn | sonne|work: yes on 'your' website ;) | 11:05 |
@lisitsyn | http://mloss.org/workshop/nips13/ | 11:05 |
sonne|work | lisitsyn: it is good that they continue this but I guess they ran out of manpower | 11:05 |
sonne|work | so if someone wants to submit shogun and go to nips... | 11:06 |
@wiking | yeah submit! | 11:06 |
@wiking | who pays? :) | 11:07 |
sonne|work | you yourself or someone who funds you :/ | 11:08 |
@lisitsyn | I would like but I would be fired | 11:08 |
@lisitsyn | :D | 11:08 |
@wiking | /Users/wiking/shogun/src/interfaces/cmdline_static/CmdLineInterface.h:7:10: fatal error: 'ui/SGInterface.h' file not found | 11:10 |
-!- dkyang [7ae07e1e@gateway/web/freenode/ip.122.224.126.30] has joined #shogun | 11:10 | |
@wiking | is there a good reason why SGInterface is being includede like this and not the standard <shogun/ui/SGInterface.h> ? | 11:10 |
@lisitsyn | ha | 11:10 |
@lisitsyn | wiking: is ithe same in other interfaces? | 11:11 |
@wiking | dunno i'm just testing cmdline | 11:11 |
sonne|work | wiking: error | 11:11 |
@wiking | lisitsyn: yes indeed r_static has the same | 11:11 |
@lisitsyn | ha | 11:12 |
sonne|work | not intentional | 11:12 |
@lisitsyn | well I'd change it | 11:12 |
@lisitsyn | as sonne|work said it is not intentional please change it then ;) | 11:12 |
sonne|work | wiking: seen my reply about modshogun? | 11:12 |
@wiking | sonne|work: ah the swig modular? | 11:12 |
sonne|work | wiking: yeah | 11:13 |
@wiking | yeah i saw that one | 11:13 |
sonne|work | wiking: so you understood? | 11:14 |
@wiking | yeps | 11:15 |
@wiking | sonne|work: btw: cmdline interface: is that like the .o-s in that directory and linked to libshogun shared? | 11:16 |
sonne|work | wiking: yes | 11:16 |
@wiking | ok i 'think' it's ready | 11:17 |
@wiking | sonne|work: what was the name of the python static lib? | 11:24 |
sonne|work | wiking: sg.so | 11:25 |
sonne|work | wiking: same for the others | 11:25 |
sonne|work | wiking: except for elwms.so | 11:25 |
dkyang | Hi,I'd like to get involved in shogun.Is there any module need to be worked on? | 11:26 |
@wiking | dkyang: :) a lot :> | 11:27 |
sonne|work | dkyang: any preferences? | 11:28 |
dkyang | I don't know how to say it, sorry for my poor english. :) Maybe something like feature learning? | 11:31 |
dkyang | sparse coding? | 11:31 |
@wiking | dkyang: yeah mlpack has that one it's good to start from what they have ;) | 11:32 |
@wiking | sonne|work: status of r_modular? | 11:33 |
sonne|work | dkyang: yeah we don't have anything like that so your contributions would be welcome | 11:34 |
sonne|work | wiking: well it compiles and I was able to run some very basic stuff | 11:34 |
dkyang | Thanks for helping,I'm glad if I can do something beneficial. | 11:36 |
@wiking | sonne|work: what's with ref counting | 11:36 |
@wiking | ? | 11:36 |
@wiking | sonne|work: you've tested i guess w/o refcounting or something? | 11:36 |
@wiking | sonne|work: afaik cmake has been not really tested with r_modular :) | 11:36 |
sonne|work | wiking: well I didn't test with cmake no | 11:37 |
@wiking | sonne|work: oke... refcounting is turned of in case of r_modular? | 11:37 |
sonne|work | wiking: only with configure but I guess you did the same with R? | 11:37 |
sonne|work | wiking: I don't remember if that was necessary | 11:37 |
@wiking | sonne|work: yeah i've treid | 11:37 |
@wiking | *tried | 11:37 |
sonne|work | what did happen? | 11:38 |
@wiking | i mean i've tried to mimic what ./configure did | 11:39 |
@wiking | ok let's see now the buildbot... | 11:40 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 11:40 | |
shogun-notifier- | shogun: Viktor Gal :develop * bbbd88e / / (4 files): https://github.com/shogun-toolbox/shogun/commit/bbbd88ec2bf8c5c092a4985075a13dcae2a0e6c5 | 11:40 |
shogun-notifier- | shogun: Add cmdline_static interface | 11:40 |
shogun-notifier- | shogun: Viktor Gal :develop * 28926e9 / / (4 files): https://github.com/shogun-toolbox/shogun/commit/28926e93a9886e5aa25de49c44c0fe85c6d07674 | 11:40 |
shogun-notifier- | shogun: Add PytonStatic interface | 11:40 |
shogun-notifier- | shogun: Viktor Gal :develop * 36e8d22 / CMakeLists.txt,src/interfaces/r_static/CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/36e8d22d651b95b627b2753bb9c1815375d8140e | 11:40 |
shogun-notifier- | shogun: Add RStatic interface | 11:40 |
sonne|work | wiking: did you add taht to the static buildbot setting? | 11:40 |
@wiking | sonne|work: i'm just on it atm | 11:42 |
sonne|work | kk | 11:43 |
@wiking | sonne|work: packages? | 11:43 |
@wiking | :) | 11:43 |
@wiking | sonne|work: ipython notebooks? :) | 11:43 |
@wiking | sonne|work: new server ? :) | 11:43 |
@wiking | sonne|work: r u editin now master.cfg on 7nn? | 11:44 |
sonne|work | wiking: just edit it | 11:46 |
sonne|work | wiking: I think I left a screen session open | 11:46 |
sonne|work | wiking: we need a meeting to decide that we rent a new server | 11:47 |
sonne|work | I cannot just spend money at will :) | 11:47 |
sonne|work | wiking: and ipython well - things are there but we need to present them somewhere/somehow | 11:47 |
@wiking | sonne|work packages? | 11:47 |
sonne|work | what? | 11:47 |
@wiking | iglesiasg: lisitsyn r u against renting a server for shogun purposes from the gsoc money? :D | 11:48 |
@wiking | sonne|work: well now we have osx native package + debian tar.gz | 11:48 |
@iglesiasg | wiking, of course not | 11:48 |
@wiking | sonne|work: we should start uploading them somewhere | 11:48 |
@lisitsyn | wiking: I don't mind even if sonne|work buys a giant pony | 11:49 |
sonne|work | wiking: we should create those nightly only when all tests pass | 11:49 |
@wiking | sonne|work: there u go it's decided... rent the server :) or of course it'd be better if the fundation would actually directly rent that server :S | 11:50 |
sonne|work | lisitsyn, wiking, iglesiasg - it will be 50EUR / month for this machine http://www.hetzner.de/en/hosting/produkte_rootserver/ex40 | 11:50 |
@wiking | sonne|work: i'm totally cool with it.... i have hetzner server myself for the last 2 years and i'm totally satisfied with their service | 11:51 |
@lisitsyn | as I said - okay :) | 11:51 |
sonne|work | well yeah I write something like our shogun toolbox foundation as name on the bill | 11:51 |
@wiking | sonne|work: can we do that? | 11:51 |
@wiking | i mean before actually the fundation is settled? | 11:51 |
@iglesiasg | wow that is an awesome machine | 11:51 |
@wiking | iglesiasg: the backend is really cool as well... fast and stable | 11:52 |
sonne|work | wiking: errm setup costs 99 $?? | 11:52 |
@wiking | sonne|work: 50eur | 11:52 |
@wiking | no? | 11:52 |
@wiking | there's setup cost for sure | 11:52 |
@wiking | i know that | 11:52 |
sonne|work | 100 EUR yes | 11:52 |
@wiking | oh woah! fuck | 11:52 |
@wiking | bastards | 11:52 |
@wiking | they raised the setup cost | 11:52 |
sonne|work | http://www.ovh.de/dedicated_server/isgenug.xml | 11:53 |
sonne|work | or we take that one | 11:53 |
sonne|work | only 24GB but 36 EUR/month | 11:53 |
@wiking | 36? | 11:53 |
@wiking | mmm | 11:53 |
@wiking | sonne|work: have u read somewhere any reviews about them | 11:54 |
@wiking | ? | 11:54 |
sonne|work | wiking: well shogun-toolbox.org is running there | 11:54 |
sonne|work | for ~1 year and so are my personal servers | 11:54 |
@wiking | ah cool | 11:54 |
@wiking | then you know them | 11:54 |
sonne|work | I am happy | 11:54 |
@wiking | ok w8 | 11:55 |
@wiking | i've just got this fancy page | 11:55 |
@wiking | http://serverbear.com/ | 11:55 |
@wiking | :) | 11:55 |
@wiking | letssee what we can get there | 11:55 |
shogun-buildbot | build #1467 of deb2 - static_interfaces is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/1467 blamelist: Viktor Gal <viktor.gal@maeth.com> | 11:56 |
sonne|work | btw I did my tax decl. - so we will see soon how much shogun money we have | 11:58 |
@lisitsyn | any estimate? :D | 11:58 |
sonne|work | lisitsyn: in the worst case 50% of the money is gone | 11:59 |
@wiking | :( | 11:59 |
sonne|work | wasted to build the best airport ever BER! | 11:59 |
@lisitsyn | hahaha | 11:59 |
@wiking | :DDDD | 11:59 |
@wiking | so 36 gigs vs 24 gigs of ram | 11:59 |
@wiking | 50 vs 36 | 12:00 |
@lisitsyn | schoenefeld is ok why to build another one ;) | 12:00 |
@iglesiasg | are they doing a third one in Berlin? | 12:00 |
@wiking | that's like 168 euros/year :( | 12:00 |
sonne|work | we actually had 3 | 12:00 |
@lisitsyn | iglesiasg: close these two and open one big | 12:00 |
@iglesiasg | lisitsyn, aham | 12:00 |
sonne|work | wiking: 362 / year | 12:00 |
@wiking | sonne|work: i mean the diff | 12:00 |
sonne|work | ahh k | 12:00 |
shogun-buildbot | build #1498 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1498 blamelist: Viktor Gal <viktor.gal@maeth.com> | 12:01 |
thoralf | sonne|work, wiking: What are the minimal requirements for the build bot? | 12:01 |
@wiking | well we could use that extra 12 gigs of ram :) | 12:01 |
@wiking | thoralf: depends | 12:01 |
@wiking | thoralf: if u dont need modular interface | 12:01 |
@wiking | then it's like 1.5 gigs ram :P | 12:02 |
@lisitsyn | guys we've got to reduce our memory/CPU requirements | 12:02 |
@lisitsyn | I am scaried we need 432432 GB to build it | 12:02 |
@lisitsyn | :) | 12:02 |
@wiking | lisitsyn: heheh yeah i couldn't compile shogun on DO | 12:02 |
@wiking | even with a 2gigs vsp | 12:02 |
@wiking | *vps | 12:02 |
@wiking | lisitsyn: and i didn't even tried to compile modular | 12:02 |
@wiking | only the simple libshogun | 12:03 |
sonne|work | lunchtime | 12:03 |
@lisitsyn | sounds like a good idea | 12:03 |
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Quit: buildmaster reconfigured: bot disconnecting] | 12:03 | |
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun | 12:03 | |
@wiking | it's only 12 :P | 12:03 |
@wiking | shogun-buildbot: force build branch=--develop 'deb2 - static_interfaces' | 12:04 |
shogun-buildbot | no such builder 'branch=--develop' | 12:04 |
@lisitsyn | wiking: 14 | 12:04 |
@lisitsyn | ;) | 12:04 |
@wiking | shogun-buildbot: force build --branch=develop 'deb2 - static_interfaces' | 12:04 |
-!- dkyang [7ae07e1e@gateway/web/freenode/ip.122.224.126.30] has quit [Quit: Page closed] | 12:04 | |
shogun-buildbot | The build has been queued, I'll give a shout when it starts | 12:04 |
@lisitsyn | wiking: sonne|work wakes up like 3 am I guess :D | 12:05 |
@wiking | ok my hadoop cluster is back | 12:12 |
@wiking | time to test the hadoop + shogun | 12:12 |
@wiking | oh shit | 12:13 |
@wiking | no the other nodes are down :( | 12:13 |
shogun-buildbot | build #1288 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1288 blamelist: Viktor Gal <viktor.gal@maeth.com> | 12:23 |
shogun-buildbot | build forced [ETA 7m01s] | 12:31 |
shogun-buildbot | I'll give a shout when the build finishes | 12:31 |
shogun-notifier- | shogun: Viktor Gal :develop * f1050f3 / CMakeLists.txt/ (3 files): https://github.com/shogun-toolbox/shogun/commit/f1050f33f0ca2836651ed2f369459d9457ce4e07 | 12:33 |
shogun-notifier- | shogun: Add OctaveStatic interface | 12:33 |
shogun-notifier- | shogun: fix RStatic library | 12:33 |
-!- travis-ci [~travis-ci@ec2-23-20-245-72.compute-1.amazonaws.com] has joined #shogun | 12:35 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/10529730 | 12:35 |
-!- travis-ci [~travis-ci@ec2-23-20-245-72.compute-1.amazonaws.com] has left #shogun [] | 12:35 | |
@wiking | mmm what has happend :S | 12:36 |
@wiking | ah ok it went actually green :D | 12:36 |
shogun-buildbot | build #1499 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1499 blamelist: Viktor Gal <viktor.gal@maeth.com> | 12:37 |
@wiking | time to finish mahout vector reader/writer | 12:42 |
shogun-notifier- | shogun: Thoralf Klein :develop * f7cb4a3 / src/shogun/evaluation/ (2 files): https://github.com/shogun-toolbox/shogun/commit/f7cb4a3a028b2f6d8ba4b94ee9b3f7df3602b87e | 12:43 |
shogun-notifier- | shogun: Using the same CRandom instance over multiple threads is not safe. Modified | 12:43 |
shogun-notifier- | shogun: StratifiedCrossValidationSplitting so that we can set own RNGs using member | 12:43 |
shogun-notifier- | shogun: m_rng. | 12:43 |
shogun-notifier- | shogun: Thoralf Klein :develop * 0cb8103 / src/shogun/lib/SGVector.cpp,src/shogun/lib/SGVector.h: https://github.com/shogun-toolbox/shogun/commit/0cb810389a1f0291bc0ac84a6a0a676fc5bb44b0 | 12:43 |
shogun-notifier- | shogun: Added additional permute methods to SGVector which take external random state. | 12:43 |
shogun-notifier- | shogun: Thoralf Klein :develop * 635a621 / src/shogun/evaluation/ (3 files): https://github.com/shogun-toolbox/shogun/commit/635a621e0453fda48e8d1570821af0ceef488b1e | 12:43 |
shogun-notifier- | shogun: Added external random state changes to CrossValidationSplitting as well. | 12:43 |
shogun-notifier- | shogun: Thoralf Klein :develop * ff6763d / src/shogun/ (3 files): https://github.com/shogun-toolbox/shogun/commit/ff6763d4eeeda1acdbc7566d34c8d8b8006d28e1 | 12:43 |
shogun-notifier- | shogun: Added shuffle(CRandom*) methods to DynArray, DynamicArray and DynamicObjectArray for thread safe shuffling. (e.g. parallel cross validation) | 12:43 |
shogun-notifier- | shogun: Thoralf Klein :develop * 33656c7 / src/shogun/evaluation/ (2 files): https://github.com/shogun-toolbox/shogun/commit/33656c7d45ce9600dff9520a92ba91884e953ae2 | 12:43 |
shogun-notifier- | shogun: Cleaned code duplication introduced by earlier commits. Now using ->shuffle(m_rng) for shuffling arrays. | 12:43 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 021393b / src/shogun/ (9 files): https://github.com/shogun-toolbox/shogun/commit/021393bce3d45f1515c4e12467259d4a742d18c2 | 12:43 |
shogun-notifier- | shogun: Merge pull request #1424 from tklein23/thread_safe_xvalidation | 12:43 |
shogun-notifier- | shogun: | 12:43 |
shogun-notifier- | shogun: Setting RNG in StratifiedCrossValidationSplitting for thread safety | 12:43 |
@wiking | sonne|work: i think we'd need this for the packages: http://docs.buildbot.net/0.8.0/full.html#Dependent-Scheduler | 12:51 |
sonne|work | wiking: I would rather not do it extra - since then we need to compile again | 12:52 |
sonne|work | wiking: but yes we have this to only compile libshogun first - when that works compile the other interfaces | 12:54 |
shogun-buildbot | build #1582 of deb3 - modular_interfaces is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1582 | 12:54 |
shogun-buildbot | build #1289 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1289 blamelist: Viktor Gal <viktor.gal@maeth.com> | 12:58 |
shogun-buildbot | build #1469 of deb2 - static_interfaces is complete: Failure [failed compile cmdline_static] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/1469 blamelist: Viktor Gal <viktor.gal@maeth.com> | 13:03 |
@wiking | sonne|work: mmmm but let's say in case of osx package. if it compiles right and the test are 100% then it should just create the package regardless what's the status of cygwin or linux bots, or? | 13:05 |
shogun-buildbot | build #1468 of deb2 - static_interfaces is complete: Failure [failed compile cmdline_static] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/1468 | 13:06 |
-!- HeikoS [~heiko@nat-176-108.internal.eduroam.ucl.ac.uk] has joined #shogun | 13:07 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 13:07 | |
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Quit: buildmaster reconfigured: bot disconnecting] | 13:08 | |
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun | 13:08 | |
@wiking | HeikoS: yo | 13:08 |
@HeikoS | wiking: hi! | 13:08 |
@HeikoS | dude | 13:08 |
@HeikoS | never ever upgrade ubuntu | 13:08 |
@HeikoS | evil | 13:08 |
@wiking | :D | 13:08 |
@wiking | hahahah | 13:08 |
@wiking | that was alwaaaays like that | 13:08 |
@wiking | since forever | 13:08 |
@HeikoS | wiking: man its 2013 | 13:08 |
@HeikoS | mankind is flying to the moon and stuff | 13:08 |
@HeikoS | I can expect that my computer still works after an upgrade | 13:09 |
@HeikoS | haha :) | 13:09 |
@HeikoS | anyway, how is it going? | 13:09 |
@HeikoS | ah sweet, virtual desktops got deactivated | 13:09 |
shogun-notifier- | shogun: Viktor Gal :develop * 521cc4b / src/interfaces/cmdline_static/CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/521cc4b40844034629fe136fc9e9dd32b9869cd0 | 13:09 |
shogun-notifier- | shogun: Rename cmdline_static target | 13:09 |
@wiking | HeikoS: well i'm just thinking about finishing up the automated tests | 13:10 |
@wiking | but quite frankly | 13:10 |
@wiking | in case of clone, equals, clone+equals and get_name | 13:10 |
@wiking | the python script would be exactly the same | 13:10 |
@HeikoS | wiking: yes | 13:11 |
@HeikoS | but if they are executed one after another | 13:11 |
@HeikoS | that helps tracking down the problem | 13:11 |
@HeikoS | many people asked me why their clone test fails | 13:11 |
@HeikoS | and it was always something get_name | 13:12 |
@HeikoS | or non-initialised memory | 13:12 |
@HeikoS | so just creating empty instance is useful | 13:12 |
@wiking | yeah yeah | 13:12 |
@HeikoS | get name also | 13:12 |
@wiking | no i was just thinking about the script itself | 13:12 |
@HeikoS | ah I see | 13:12 |
@wiking | that i shoudl make it in a way | 13:12 |
@wiking | that they share some parts | 13:12 |
@HeikoS | wiking: well maybe just accept the duplicate code | 13:12 |
@wiking | and not copy-paste | 13:12 |
@HeikoS | dont know whether thats worth it | 13:12 |
@wiking | HeikoS: well it's just putting the stuff into a common .py | 13:13 |
@wiking | and then import that into all the scripts | 13:13 |
@HeikoS | wiking: if its not too much hassle, then maybe | 13:13 |
@wiking | yeah i'll do that now | 13:13 |
@HeikoS | but also makes it harder to undetrstand for new devs | 13:13 |
thoralf | wiking: Is there a make target which only builds the unit-tests but does not execute them? | 13:13 |
@wiking | thoralf: make shogun-unit-test | 13:13 |
@wiking | that creates the exec | 13:13 |
thoralf | wiking: Cool, thanks. | 13:13 |
@wiking | HeikoS: well it's python script :P | 13:13 |
@wiking | HeikoS: i u dont understand what's import you'll suck at it anyways | 13:14 |
@HeikoS | wiking: do you have any idea why a webbrowser opens everytime I clock on the sidebar | 13:14 |
@wiking | eh? :D | 13:14 |
@HeikoS | wiking: well then go ahead :) | 13:14 |
@wiking | HeikoS: i'm more concerned if i could set the order | 13:15 |
@wiking | of the tests | 13:15 |
@HeikoS | wiking: yes thats important | 13:15 |
@wiking | as u said it'd be great to have: get_name, clone, clone+equals | 13:15 |
@wiking | but i dont know if i can do that with gtest | 13:15 |
@wiking | HeikoS: "By default, Google Test uses a random seed calculated from the current time. Therefore you'll get a different order every time." | 13:18 |
@wiking | HeikoS: so basically if we want to have the order | 13:18 |
@wiking | we do the whole test within one block | 13:18 |
@wiking | :S | 13:18 |
@wiking | i mean you'll get the line number anyways where did it fail | 13:19 |
@HeikoS | wiking: mmmmh | 13:19 |
@HeikoS | what about allowing them to fail? | 13:19 |
@HeikoS | then if say get_name is wrong | 13:19 |
@HeikoS | all of them fail | 13:19 |
@HeikoS | providing quite some information | 13:19 |
@HeikoS | if there is a bug in clone, only that last one will fail | 13:19 |
@HeikoS | so the order is in fact not important | 13:20 |
@wiking | ? | 13:20 |
@wiking | allow them to fail? | 13:20 |
@wiking | i mean in gtest if a test fail (except segfault) it'll run all the tests | 13:20 |
@wiking | *fails | 13:20 |
@wiking | but of course you'll have 3 fails because of get_name() | 13:20 |
@wiking | HeikoS: why dont we do this in one codeblock? | 13:21 |
@HeikoS | wiking: I dont like one codeblock | 13:21 |
@wiking | i mean if a getname fails | 13:22 |
@HeikoS | hiding infos | 13:22 |
@wiking | then it won't try to run clone and equals | 13:22 |
@HeikoS | or what do you mean? | 13:22 |
@HeikoS | wiking: is that clearly visible then? | 13:22 |
@HeikoS | if yes, then its fine | 13:22 |
@wiking | HeikoS: yes yes | 13:22 |
@HeikoS | wiking: another problem is testing for memory | 13:22 |
@HeikoS | is it possible to seperately test all of them? | 13:22 |
@wiking | HeikoS: it'll tell you exactly which test (which line) failed exactly | 13:22 |
@HeikoS | if we later on make valgrind mandatory? | 13:22 |
@wiking | HeikoS: define separate :) | 13:23 |
-!- travis-ci [~travis-ci@ec2-54-224-203-225.compute-1.amazonaws.com] has joined #shogun | 13:23 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/10531634 | 13:23 |
-!- travis-ci [~travis-ci@ec2-54-224-203-225.compute-1.amazonaws.com] has left #shogun [] | 13:23 | |
@HeikoS | ok then | 13:24 |
@wiking | HeikoS: u can run WhateverClass.whicevertest separately | 13:24 |
@wiking | if u want | 13:24 |
@wiking | you can run WhateverClass.* | 13:24 |
@HeikoS | wiking: ok then we can use this for memory testing later | 13:24 |
@wiking | HeikoS: well if we put of course | 13:24 |
sonne|work | HeikoS: well use debian if you want something stable :P | 13:25 |
@wiking | getname + clone + clone equals | 13:25 |
@wiking | in one test | 13:25 |
@wiking | then of course you wont be able to separete that | 13:25 |
@wiking | HeikoS: the bigger problem is actually | 13:25 |
@wiking | that we have memchecking support with cmake | 13:25 |
@wiking | but | 13:25 |
sonne|work | wiking: wrt osx and packages - we should IMHO build for all interfaces and then create a big package from that | 13:25 |
@HeikoS | sonne|work: I know ;) I tried, but software too old, I just needed a more recent scipy which support some more sparse stuff | 13:25 |
@wiking | that would run valgrind on the whole unit-test.... | 13:25 |
@wiking | sonne|work: yeps... but only nightly... or? | 13:26 |
@wiking | sonne|work: or you want to always build on that poor osx buildbot all the interfaces :D | 13:26 |
@wiking | HeikoS: ok i'm modifying now the clone_eqauls test | 13:27 |
@HeikoS | wiking: cool! | 13:27 |
@wiking | and then i'll add serialisation test for all the backends | 13:27 |
@wiking | :) | 13:27 |
@HeikoS | wiking: and then I have to fix stuff I guess ;) | 13:27 |
@HeikoS | wiking: what do you think about the non-trivial cases? | 13:27 |
sonne|work | wiking: no I mean we should do all the doc/package/etc stuff in our nightly default - and nightly default should have all the supported interfaces enabled / to be build. so all static/modular | 13:27 |
@wiking | sonne|work: yeah i get that one | 13:28 |
@wiking | sonne|work: but we need to define nightly_default for osx buildbot as well | 13:28 |
@wiking | sonne|work: if we want to supply osx packages | 13:28 |
@wiking | HeikoS: huh | 13:28 |
sonne|work | wiking: I am using that poor osx macbook air as my travel machine for shogun now so around daytime it won't build btw but only evening | 13:28 |
@wiking | sonne|work: ehehehhe :))) | 13:28 |
sonne|work | wiking: so what do we do about the server - take the 24G one? | 13:28 |
@HeikoS | wiking: classes with content rather than everything NULL | 13:28 |
@wiking | sonne|work: i've checked licensing about osx | 13:28 |
@wiking | sonne|work: we can have osx buildbot in a VM | 13:29 |
sonne|work | wiking: well 450EUR for a dedicated osx buildbot is a bit too much no? | 13:29 |
@wiking | sonne|work: i own 2 licenses of OSX that allows me to run 4 osxes | 13:29 |
sonne|work | HeikoS: are you ok with us ordering another server for buildbot etc where we all will have logins? | 13:29 |
@wiking | sonne|work: even in a VM | 13:29 |
sonne|work | wiking: ahh ok nice | 13:29 |
@wiking | sonne|work: but but! | 13:29 |
@HeikoS | sonney2k: yes I am | 13:29 |
sonne|work | wiking: but but? | 13:30 |
@wiking | sonne|work: think about the ram that OSX VM will eat | 13:30 |
sonne|work | osx server? | 13:30 |
@wiking | no | 13:30 |
@wiking | we cannot have that :( | 13:30 |
@wiking | the license does not allow osx server | 13:30 |
@wiking | :( | 13:30 |
@wiking | only osx desktop | 13:30 |
sonne|work | wiking: I think you need quite some hacks to run osx inside a VM that is not on a mac | 13:30 |
@wiking | sonne|work: virtualbox | 13:30 |
sonne|work | wiking: IIRC you are allowed to do that only on a mac | 13:30 |
@wiking | sonne|work: apt-get install and thats allo | 13:30 |
@wiking | *all | 13:31 |
@wiking | nono | 13:31 |
@HeikoS | wiking: with the new ubuntu, I gett locally | 13:31 |
@HeikoS | make[2]: *** No rule to make target `/usr/lib/libnlopt.so', needed by `src/shogun/libshogun.so.14.0'. Stop. | 13:31 |
@wiking | i've read about this in the VB forums | 13:31 |
@wiking | and tickets | 13:31 |
sonne|work | wiking: well we give osx 8 GB - should be sufficient right? and btw we could also only run things at a time | 13:31 |
@HeikoS | the file is here | 13:31 |
@HeikoS | /usr/lib/x86_64-linux-gnu/libnlopt.so | 13:31 |
sonne|work | HeikoS: well rm -rf build | 13:31 |
@HeikoS | whose fault is that? | 13:31 |
@wiking | HeikoS: fire a bug :) | 13:31 |
@HeikoS | ah | 13:31 |
sonne|work | HeikoS: and redo cmake | 13:31 |
@HeikoS | make clean doesnt do it? | 13:31 |
@wiking | HeikoS: but yeah first try to clean the cmake | 13:31 |
@HeikoS | I did ccmake clean .. | 13:31 |
@wiking | HeikoS: rm -rf build is safe ;) | 13:32 |
@wiking | sonne|work: mmm yeah 8 gigs should be fine | 13:32 |
sonne|work | wiking: I think the 24G are sufficient or we have to order another one :D | 13:32 |
@wiking | even 4 | 13:32 |
@wiking | or lets say 6 | 13:32 |
@wiking | and then we for sure can build the modular interfaces as well | 13:32 |
@wiking | sonne|work: heheheh :D | 13:32 |
@wiking | sonne|work: so let's say base system + osx bot | 13:32 |
@wiking | what else | 13:32 |
@wiking | just to see if we fit in nicely | 13:32 |
@HeikoS | ah that did it thanks | 13:33 |
sonne|work | wiking: well the interactive demos | 13:33 |
sonne|work | wiking: they might eat a few GB | 13:33 |
sonne|work | lets say 4 | 13:33 |
@wiking | ok that's like 10 with the osx bot | 13:33 |
@wiking | so we have like 12 more | 13:33 |
@wiking | the base system will eat at least 2gigs i would say | 13:34 |
sonne|work | 24 -4 -6 == 14 ? | 13:34 |
sonne|work | wiking: nope maybe 200MB | 13:34 |
@wiking | sonne|work: u recon | 13:34 |
@wiking | reckon | 13:34 |
sonne|work | then buildbot ~200MB | 13:34 |
@wiking | ok | 13:34 |
@wiking | + nginx | 13:34 |
@wiking | i guess | 13:34 |
sonne|work | ngings 100MB | 13:34 |
sonne|work | x | 13:34 |
@wiking | do we need an sql? | 13:34 |
sonne|work | maybe not | 13:35 |
sonne|work | no idea | 13:35 |
sonne|work | but say another 500mb | 13:35 |
@wiking | sonne|work: would we move the webpage there ? | 13:35 |
sonne|work | still plenty room | 13:35 |
@wiking | ok | 13:35 |
sonne|work | wiking: well if possible I would keep the current webserver alive | 13:35 |
@wiking | we can have at least one more VM or? | 13:35 |
@wiking | i mean for testing purposes | 13:35 |
sonne|work | wiking: and let it do the nightly builds but not much more | 13:36 |
@wiking | lets say somebody comes in with a funky error | 13:36 |
@wiking | with a funky os | 13:36 |
sonne|work | wiking: yeah but why a VM? | 13:36 |
sonne|work | I mean it is rather a scratch setup | 13:36 |
@wiking | we could just run a new VM instance fast and try out the stuff | 13:36 |
sonne|work | or just try directly in a chroot | 13:36 |
@wiking | sonne|work: yeah if it's linux | 13:36 |
@wiking | sonne|work: btw how much is a win license nowadays? | 13:37 |
@wiking | lisitsyn: iglesiasg HeikoS ^ | 13:37 |
@HeikoS | ? | 13:37 |
@wiking | sonne|work: because we should start thinking about windows builds | 13:37 |
@iglesiasg | wiking, no idea | 13:37 |
@HeikoS | yes +1 | 13:37 |
@wiking | sonne|work: with cmake we can do it... | 13:37 |
sonne|work | wiking: well we have a windows machine already | 13:37 |
@HeikoS | I dont know havent used windows for quite some time | 13:38 |
@wiking | sonne|work: could we run there a pure MSVC based buildbot as well? | 13:38 |
sonne|work | HeikoS: I just used it for my tax decl. | 13:38 |
sonne|work | wiking: sure | 13:38 |
sonne|work | it has sufficient mem and disk | 13:38 |
sonne|work | just single core in the vm but that is not the problem | 13:38 |
@wiking | sonne|work: ok then when u have time plz just install buildbot + MSVS + cmake on that machine | 13:38 |
sonne|work | wiking: I don't have MSVS | 13:38 |
@wiking | sonne|work: express is for free | 13:39 |
sonne|work | and I didn't get a reply from the guy at the institute | 13:39 |
sonne|work | express can build shogun? | 13:39 |
@wiking | sonne|work: http://www.microsoft.com/visualstudio/eng/products/visual-studio-express-products | 13:39 |
@wiking | sonne|work: yeah i'm working on it... but we'll get there | 13:39 |
@wiking | sonne|work: with cmake i have already the solution file for VS | 13:39 |
@wiking | sonne|work: needs some hacking but will get there... | 13:40 |
@wiking | sonne|work: so i'm just saying that one way or another we should have a pure windows based buildbot | 13:40 |
@wiking | either on this new machine or somewhere... | 13:40 |
sonne|work | wiking: which version? | 13:41 |
@wiking | sonne|work: windows or VS? :) | 13:41 |
@wiking | vs doesn't matter | 13:41 |
@wiking | would be great if it's 2010 or later | 13:41 |
sonne|work | well windows we have 7 and sry no 8 for me | 13:41 |
@wiking | ok so win7 + vs 2010 or later | 13:41 |
sonne|work | and we need to sign up w/ M$ | 13:42 |
@wiking | cmake will generate the right solution file... | 13:42 |
sonne|work | HeikoS: so about the trip - I am planning to go on the 18th and then stay the week after (to be back on the weekend) | 13:42 |
sonne|work | HeikoS: so rent a car etc | 13:42 |
@wiking | i mean if this guy who owns the cygwin buildbot can setup a VS env for us | 13:42 |
@wiking | we are set | 13:42 |
sonne|work | HeikoS: we could meet Olivier/Quoc or so ... | 13:43 |
@wiking | if not then we should really think about getting a windows license :S | 13:43 |
sonne|work | wiking: well I administer this thing so sure we can do | 13:43 |
@wiking | sonne|work: kayak.com last time i was there i've got a ford mustang cabrio for 15USD/day | 13:43 |
sonne|work | wiking: and it has a windows license | 13:43 |
@HeikoS | sonne|work: so go back on 25? | 13:43 |
@HeikoS | sonne|work: ok I am fine with that! | 13:43 |
sonne|work | yeah around that | 13:43 |
@HeikoS | sonne|work: should be book flights together? | 13:44 |
@wiking | sonne|work: ok cool so then just install the stuff on that machine. native python+cmake+vs express+buildbot | 13:44 |
@HeikoS | when should we book? | 13:44 |
@HeikoS | soon or wait for google money? | 13:44 |
@HeikoS | sonne|work: where do you want to stay? We could rent a small flat on airbnb or so | 13:44 |
@wiking | sonne|work: see again c# compiled fine on slave4: http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1583 | 13:45 |
@wiking | :S | 13:45 |
sonne|work | HeikoS: I was planning to be at the summit from 18-20 and then still sleep in the hotel for 1 night | 13:45 |
@HeikoS | sonne|work: and afterwards? | 13:46 |
sonne|work | then e.g. meet people at google or so and then go to santa cruz | 13:46 |
@HeikoS | I mean sleeping | 13:46 |
sonne|work | maybe some national park (and hike there) | 13:46 |
@HeikoS | yeah totally | 13:46 |
@HeikoS | so stay somewhere else every night? | 13:46 |
@wiking | sonne|work: go here, it's really great place to hang around: http://en.wikipedia.org/wiki/17-Mile_Drive | 13:47 |
@HeikoS | I think I am pretty much open to everything :) | 13:48 |
@wiking | sonne|work: ok so if we dont need that windows buildbot extra then we should just get that machine for 36 euros/month | 13:48 |
@wiking | HeikoS: https://gist.github.com/vigsterkr/6318518 | 13:55 |
@wiking | HeikoS: would u b ok with this? | 13:55 |
@HeikoS | wiking: yes! but do get name first | 13:55 |
@wiking | HeikoS: isn't it the first? :) | 13:56 |
@wiking | https://gist.github.com/vigsterkr/6318518#file-clone_unittest-cc-jinja2-L8 | 13:56 |
@HeikoS | ah sorry yes | 13:56 |
@wiking | the first assert is really just about getting the object ;P | 13:56 |
@wiking | ok i'll do this for the temp classes | 13:56 |
@wiking | and here we go | 13:56 |
@HeikoS | object2 has to be unrefeed | 13:56 |
shogun-buildbot | build #1500 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1500 blamelist: Viktor Gal <viktor.gal@maeth.com> | 13:57 |
@HeikoS | ah it is | 13:57 |
@HeikoS | sorry | 13:57 |
@HeikoS | no its good :) | 13:57 |
shogun-buildbot | build #1470 of deb2 - static_interfaces is complete: Failure [failed compile octave_static] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/1470 blamelist: Viktor Gal <viktor.gal@maeth.com> | 13:57 |
@HeikoS | hopefully this wont fail ;) | 13:57 |
@wiking | ok i'll test it first locally :D | 13:58 |
shogun-buildbot | build #1471 of deb2 - static_interfaces is complete: Failure [failed compile cmdline_static] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/1471 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Thoralf Klein <thoralf.klein@zib.de>, Viktor Gal <viktor.gal@maeth.com> | 13:59 |
@wiking | eh why not buildbot?! | 13:59 |
@wiking | HeikoS: ok so what about obtain from generic? | 14:00 |
@wiking | HeikoS: while we are at hacking jinja2 templates :D | 14:00 |
@HeikoS | wiking: since devs can easily use dynamic cast with an assert (they KNOW they expected type), we dont need this inside shogun code, so no ifdefs in there | 14:01 |
@HeikoS | so I would say lets do it optional | 14:01 |
@HeikoS | to not have the dependency | 14:01 |
@HeikoS | its mainly for modular interface users who cannot cast otherwise | 14:02 |
shogun-buildbot | build #1501 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1501 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Thoralf Klein <thoralf.klein@zib.de>, Viktor Gal <viktor.gal@maeth.com> | 14:02 |
@HeikoS | my optinion | 14:02 |
@wiking | HeikoS: okey | 14:03 |
@HeikoS | ask lisitsyn and sonne|work though | 14:03 |
@wiking | so then 1) serialization unit tests + 2) obtain from generic :) | 14:03 |
@HeikoS | cool! | 14:03 |
@HeikoS | good hacking :) | 14:03 |
@wiking | HeikoS: well it could be usefull in case of serialization fw | 14:03 |
@HeikoS | wiking: it will be | 14:04 |
@lisitsyn | yes? what? | 14:04 |
@HeikoS | wiking: the only thing is the non-trivial data | 14:04 |
@wiking | HeikoS: yeah in kow | 14:04 |
@wiking | *i know | 14:04 |
@HeikoS | that is yet unsolved and this will massively increase the usefulness of these tests | 14:04 |
@HeikoS | lisitsyn: obtain_from_generic optional | 14:04 |
@wiking | indeed | 14:04 |
@HeikoS | wiking: let me know your ideas on this :) | 14:04 |
@wiking | HeikoS: ppppfff | 14:04 |
@iglesiasg | lisitsyn, wiking, HeikoS, sonney2k do you guys know about this CombinedKernel weights issue asked in the mailing list? | 14:04 |
@wiking | HeikoS: braindead about this | 14:05 |
@lisitsyn | HeikoS: why/ | 14:05 |
@HeikoS | iglesiasg: yes, but no idea, he is right ;) | 14:05 |
@iglesiasg | in MKL the kernel weights should add up to 1, I think | 14:05 |
@iglesiasg | is combined kernel implementing that? | 14:05 |
@HeikoS | wiking: two alternating approaches | 14:05 |
@lisitsyn | HeikoS: why to make it optional? | 14:05 |
@lisitsyn | iglesiasg: no MKL* | 14:05 |
@HeikoS | iglesiasg: combined kernel can have anything | 14:05 |
@HeikoS | iglesiasg: I bet its a bug | 14:05 |
@wiking | iglesiasg: afaik anything | 14:05 |
@HeikoS | but no idea about this stuff | 14:05 |
@iglesiasg | mm | 14:06 |
@iglesiasg | so then they do not have to add up to 1 in combined kernel? | 14:06 |
@HeikoS | lisitsyn: because otherwise libshogun depends on jinja2 | 14:06 |
@wiking | iglesiasg: mmm i dont know in case of MKL | 14:06 |
@wiking | iglesiasg: but other than that the weights can be anything | 14:06 |
@HeikoS | iglesiasg: no, but usually the weights are convex | 14:06 |
@HeikoS | but I dont know, sonne|work wrote this | 14:06 |
@lisitsyn | HeikoS: ahh | 14:06 |
@iglesiasg | HeikoS, all right. Let's ask sonne|work once he is back then :) | 14:07 |
@HeikoS | wiking, iglesiasg the weights have to be bounded otherwise one cannot optimise | 14:07 |
@HeikoS | iglesiasg: yep I think thats best | 14:07 |
@iglesiasg | HeikoS, that makes sense | 14:07 |
@wiking | :D | 14:07 |
@iglesiasg | but I think it can fine if they don't add up to one | 14:07 |
@iglesiasg | but I am not sure | 14:07 |
thoralf | iglesiasg, HeikoS: I think it depends what you're learning. You can also constrain |w|_p = 1 | 14:09 |
@HeikoS | iglesiasg: it depends on what is done with them after learning | 14:09 |
thoralf | for instance 2-norm | 14:09 |
thoralf | or better |w|_p^p = 1 | 14:09 |
thoralf | don't know the exact definition | 14:10 |
@iglesiasg | thoralf, aham that makes sense too. It might be that it is another norm than L1 the one that is set to one | 14:10 |
@wiking | HeikoS: another thing... computationengine | 14:10 |
@wiking | HeikoS: i get that that you want to use mpi | 14:10 |
@wiking | HeikoS: but i seriously think that sending data over sockets is really not the most optimal stuff... we should somehow support to pass around references to data on the fs to use | 14:11 |
@wiking | because now basically what you'll do with an IndependentJob is that you add everything into it | 14:11 |
@wiking | and then serialise/deserialise it | 14:11 |
thoralf | iglesiasg: Yes. If individual weights are within [0;1], then it doesn't look that bad. ;) | 14:11 |
@wiking | and run compute in it | 14:11 |
@HeikoS | wiking: ok | 14:11 |
@HeikoS | wiking: in this moment I am writing something like this in python | 14:12 |
@HeikoS | for PBS | 14:12 |
@wiking | but i dont think it's really good idea to pass around vectors/features/matrixes | 14:12 |
@HeikoS | where one can only communicate over FS | 14:12 |
@HeikoS | wiking: Ill share the code once its done | 14:12 |
@HeikoS | almost the same framework as in shogun | 14:12 |
@HeikoS | and I will use/test it soon | 14:12 |
@HeikoS | if it works nice, we can think of adding it to shogun | 14:12 |
@HeikoS | I basically store the jobs on the FS | 14:13 |
@HeikoS | and shared data is loaded by the job | 14:13 |
@HeikoS | and the engine then checks whether certain result files exist | 14:13 |
@wiking | HeikoS: i'm just now trying to write the hadoop backend with shogun | 14:13 |
@HeikoS | and submits jobs to the PBS queue | 14:13 |
@HeikoS | wiking: wow ! | 14:13 |
@wiking | HeikoS: and there it's really you pass the job ... but not the data itself | 14:14 |
@wiking | HeikoS: as the data is sitting on the hdfs | 14:14 |
@wiking | which is obviously accessible by all the nodes | 14:14 |
@HeikoS | wiking: yes | 14:14 |
@HeikoS | well the job is a fiule | 14:14 |
@HeikoS | file | 14:14 |
-!- gsomix__ [~gsomix@95.67.189.148] has joined #shogun | 14:14 | |
@wiking | yeah i know... but then always we serialise the data | 14:14 |
@wiking | so let's say we have we have the data sitting on the fs already | 14:15 |
@wiking | a CFeatures serialised | 14:15 |
@wiking | one should be able to just say ok here's the job and that's the data there | 14:15 |
@wiking | do your job | 14:15 |
@HeikoS | wiking: I thinks this is task of the job type | 14:15 |
@HeikoS | yes the job just is executed | 14:15 |
@HeikoS | it can do this then | 14:15 |
@wiking | w8 | 14:16 |
@wiking | i'll get back to u in a sec | 14:16 |
@wiking | here's a nice one | 14:16 |
@wiking | (the unit tests came back ) | 14:16 |
@wiking | :D | 14:16 |
shogun-buildbot | build #1290 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1290 blamelist: Viktor Gal <viktor.gal@maeth.com> | 14:16 |
@wiking | just creating the pastebin | 14:16 |
-!- gsomix_ [~gsomix@178.45.75.171] has quit [Ping timeout: 248 seconds] | 14:17 | |
@HeikoS | wiking: thats exciting stuff | 14:17 |
@HeikoS | if we had a PBS backend | 14:17 |
@HeikoS | I think everyone would freak out ;) | 14:17 |
@wiking | ok | 14:19 |
@wiking | http://pastebin.com/J6GDhtVu | 14:19 |
@wiking | time to fix | 14:19 |
@wiking | :) | 14:19 |
@wiking | CStringFileFeatures does not have get_name defined... | 14:20 |
@wiking | same with nearestcentroid | 14:21 |
@wiking | ok i'll do these fixes | 14:21 |
@wiking | is it possible that somewhere intentionally we dont have a get_name() | 14:21 |
@wiking | ping iglesiasg HeikoS sonne|work lisitsyn ^ | 14:23 |
@iglesiasg | wiking, no that I know of | 14:23 |
@iglesiasg | I don't quite find a reason why either | 14:24 |
-!- gsomix__ [~gsomix@95.67.189.148] has quit [Ping timeout: 246 seconds] | 14:24 | |
@HeikoS | wiking: no | 14:24 |
@HeikoS | get_name must always be there | 14:24 |
@wiking | okok then i'm fixing the get_name errors | 14:24 |
@HeikoS | wiking: thanks! | 14:25 |
@lisitsyn | wiking: it is pure virtual | 14:28 |
@lisitsyn | but with inheritance.. | 14:29 |
@wiking | lisitsyn: heheh yeah but... indeed | 14:29 |
@wiking | :) | 14:29 |
@wiking | ok these are fixed | 14:30 |
@wiking | HeikoS: you'll have some fun now with fixing .equals | 14:30 |
@wiking | :( | 14:30 |
@wiking | should i commit the test, or we dont want constantly failing travis? | 14:31 |
@HeikoS | wiking: on empty instances? | 14:31 |
@wiking | HeikoS: yeah | 14:31 |
@HeikoS | wiking: commit it and blacklist | 14:31 |
@HeikoS | easiest for me | 14:31 |
@wiking | HeikoS: i cannot blacklist it as it's templated :P | 14:31 |
@HeikoS | wiking: thats most likely uninitialised memory, Ill have a look | 14:31 |
@HeikoS | ah | 14:31 |
@HeikoS | so how can I access? | 14:31 |
@wiking | HeikoS: i mean i can but it's really fucking awkward to do | 14:31 |
@wiking | HeikoS: ok i'll push the fix for the get_names | 14:32 |
@HeikoS | wiking: can you give me a list of classnames that fail? | 14:32 |
@HeikoS | then Ill write a program locally | 14:32 |
@wiking | and then i'll just give u the patch | 14:32 |
@wiking | and u can run locally the unit test | 14:32 |
@wiking | HeikoS: | 14:32 |
@wiking | [ FAILED ] SGObject.clone_equals_FeatureBlockLogisticRegression | 14:32 |
@wiking | [ FAILED ] SGObject.clone_equals_GaussianBlobsDataGenerator | 14:32 |
@wiking | [ FAILED ] SGObject.clone_equals_MCLDA | 14:32 |
@wiking | [ FAILED ] SGObject.clone_equals_DualLibQPBMSOSVM | 14:32 |
@wiking | [ FAILED ] SGObject.clone_equals_Factor | 14:32 |
@wiking | [ FAILED ] SGObject.clone_equals_MAPInference | 14:33 |
@wiking | [ FAILED ] SGObject.clone_equals_StringFeatures_PT_INT16 | 14:33 |
@wiking | [ FAILED ] SGObject.clone_equals_StringFeatures_PT_UINT16 | 14:33 |
@wiking | [ FAILED ] SGObject.clone_equals_StringFeatures_PT_FLOAT32 | 14:33 |
@wiking | [ FAILED ] SGObject.clone_equals_StringFeatures_PT_FLOAT64 | 14:33 |
@wiking | [ FAILED ] SGObject.clone_equals_StringFileFeatures_PT_INT16 | 14:33 |
@wiking | all failed on equals | 14:33 |
@HeikoS | ok thanks | 14:33 |
@HeikoS | will check soon | 14:33 |
@HeikoS | but now | 14:33 |
@HeikoS | lunch :) | 14:33 |
shogun-notifier- | shogun: Viktor Gal :develop * 0b0f529 / src/shogun/ (4 files): https://github.com/shogun-toolbox/shogun/commit/0b0f529cde3630b6339eb6f0974a496d0573d3fb | 14:34 |
shogun-notifier- | shogun: Add or fix get_name function for the classes in the patch | 14:34 |
@wiking | HeikoS: sent u the patch | 14:36 |
@wiking | ok now let's see about the serialization tests :P | 14:36 |
-!- gsomix__ [~gsomix@88.200.242.148] has joined #shogun | 14:37 | |
-!- gsomix__ [~gsomix@88.200.242.148] has quit [Ping timeout: 240 seconds] | 14:41 | |
shogun-buildbot | build #1291 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1291 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Thoralf Klein <thoralf.klein@zib.de>, Viktor Gal <viktor.gal@maeth.com> | 14:44 |
@wiking | btw do we agree that in gtest we should only print out failed errors? | 14:48 |
@wiking | as it's getting really a looong output | 14:48 |
-!- gsomix__ [~gsomix@95.67.172.244] has joined #shogun | 14:55 | |
@iglesiasg | wiking, I admit I like to run the binary shogun unit test by hand and see all the green :) | 14:56 |
-!- iglesiasg_ [~iglesias@n132-p33.kthopen.kth.se] has joined #shogun | 15:01 | |
-!- iglesiasg [~iglesias@2001:6b0:1:1041:5158:760a:da84:6c51] has quit [Read error: Connection reset by peer] | 15:01 | |
shogun-buildbot | build #1472 of deb2 - static_interfaces is complete: Failure [failed compile octave_static] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/1472 blamelist: Viktor Gal <viktor.gal@maeth.com> | 15:07 |
shogun-buildbot | build #1502 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1502 blamelist: Viktor Gal <viktor.gal@maeth.com> | 15:09 |
@lisitsyn | iglesiasg_: sounds like coming out | 15:12 |
@lisitsyn | :D | 15:12 |
iglesiasg_ | lisitsyn, haha yeah the same thing lol | 15:13 |
@lisitsyn | iglesiasg_: have you heard bradley manning is chelsea manning now btw? | 15:14 |
@wiking | iglesiasg_: ok i'll add it as cmdline args just for u | 15:14 |
@lisitsyn | :D | 15:14 |
@wiking | iglesiasg_: and by default it'll print everything :> | 15:14 |
iglesiasg_ | lisitsyn, yeah I saw something about that | 15:14 |
iglesiasg_ | wiking, wow it sounds like Christmas! | 15:15 |
thoralf | iglesiasg_: Yeah :D | 15:21 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.] | 15:26 | |
shogun-notifier- | shogun: Viktor Gal :develop * 3336b14 / tests/unit/base/main_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/3336b147241ff69a434f69f8b1c7e1a45e5d56a6 | 15:28 |
shogun-notifier- | shogun: Add --only-on-failure command line argument for unit tests | 15:28 |
shogun-notifier- | shogun: With --only-on-failure only failing unit tests are printed and the unit | 15:28 |
shogun-notifier- | shogun: test summary on the end. By default it's disabled. | 15:28 |
@wiking | iglesiasg_: ^ voila | 15:28 |
iglesiasg_ | viva! | 15:29 |
@wiking | ok now serialisation tets | 15:29 |
@wiking | *tests | 15:29 |
-!- gsomix__ is now known as gsomix | 15:30 | |
shogun-buildbot | build #1292 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1292 blamelist: Viktor Gal <viktor.gal@maeth.com> | 15:33 |
@wiking | HeikoS: hahah you'll be shocked about this :) | 15:45 |
@wiking | HeikoS: http://pastebin.com/iR1Wmn1y | 15:45 |
@wiking | i don't understand that how's that possible that SGObject.clone_equals errors are not a true subset of SerializationAscii | 15:46 |
@wiking | since SerializationAscii uses object->equals() | 15:46 |
@HeikoS | ? | 15:46 |
@HeikoS | the same classes fail? | 15:46 |
@wiking | no | 15:46 |
@wiking | that's the funky part | 15:46 |
@wiking | as u can see the two sets does have an intersection | 15:46 |
@wiking | but for example MCLDA doesn't fail with serialization+equals | 15:47 |
@wiking | :DDD | 15:47 |
@wiking | HeikoS: even more... 2 consequtive tests does not give the same errors with serialization :P | 15:53 |
@HeikoS | weiiiird :) | 15:54 |
@wiking | yeps | 15:54 |
@wiking | anyhow i'll write these tests | 15:54 |
@wiking | maybe i'll add it to a separate branch | 15:54 |
@wiking | so that until we dont fix it | 15:54 |
@wiking | we dont add it to the develop branch | 15:54 |
@wiking | ok? | 15:54 |
@HeikoS | thats good | 15:57 |
@HeikoS | different branch | 15:58 |
sonne|work | wiking: well ascii is lossy | 15:58 |
@wiking | sonne|work: that's ok i'm writing json hdf5 and xml as well | 15:58 |
shogun-buildbot | build #1473 of deb2 - static_interfaces is complete: Failure [failed compile octave_static] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/1473 blamelist: Viktor Gal <viktor.gal@maeth.com> | 15:59 |
shogun-buildbot | build #1503 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1503 blamelist: Viktor Gal <viktor.gal@maeth.com> | 16:01 |
-!- iglesiasg_ [~iglesias@n132-p33.kthopen.kth.se] has quit [Ping timeout: 264 seconds] | 16:08 | |
shogun-buildbot | build #1584 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1584 blamelist: Viktor Gal <viktor.gal@maeth.com> | 16:09 |
sonne|work | wiking: json is even more lossy :D | 16:14 |
@wiking | well we can fix those i hope | 16:16 |
sonne|work | wiking: no not possible | 16:19 |
sonne|work | wiking: when you have a double -> json or -> ascii | 16:19 |
sonne|work | it is always lossy | 16:19 |
@wiking | sonne|work: then why do we have it at the first place? | 16:19 |
sonne|work | errm?? | 16:20 |
sonne|work | you want to store your model as .json / ascii sometimes as hdf5 etc | 16:20 |
@wiking | well if we know that it's lossy, hence it cannot really truly serialise a class then we should just not have it | 16:20 |
sonne|work | YMMV | 16:20 |
sonne|work | why not? | 16:20 |
@wiking | well why yes? | 16:20 |
sonne|work | it is 'good enough' | 16:20 |
-!- iglesiasg_ [~iglesias@2001:6b0:1:1041:5158:760a:da84:6c51] has joined #shogun | 16:20 | |
sonne|work | people use .csv's to store data for decades and still do | 16:21 |
sonne|work | even though it is always lossy | 16:21 |
sonne|work | (for double)! | 16:21 |
shogun-buildbot | build #1293 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1293 blamelist: Viktor Gal <viktor.gal@maeth.com> | 16:25 |
-!- sonne|work [~sonnenbu@91-64-72-127-dynip.superkabel.de] has quit [Quit: Leaving.] | 16:33 | |
-!- pickle27 [~kevin@d67-193-243-174.home3.cgocable.net] has joined #shogun | 16:38 | |
-!- sonne_ [~sonne@82.113.99.43] has joined #shogun | 16:48 | |
-!- sonne_ is now known as sonne|osx | 16:51 | |
@wiking | hohooooo | 16:55 |
@wiking | segfault with XML | 16:55 |
-!- HeikoS [~heiko@nat-176-108.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 240 seconds] | 16:55 | |
shogun-notifier- | shogun: Viktor Gal :feature/SerialUTests * 45ad889 / tests/unit/ (5 files): https://github.com/shogun-toolbox/shogun/commit/45ad889864df9ae6b10627b805bd8b30a59e1aa4 | 16:56 |
shogun-notifier- | shogun: Refactor autogenerated clone_unittest | 16:56 |
shogun-notifier- | shogun: Add autogenerated unit tests for all the supported serialisation | 16:56 |
@wiking | oookeeey | 16:56 |
shogun-notifier- | shogun: backends | 16:56 |
shogun-notifier- | shogun: Viktor Gal :feature/SerialUTests * 1e9118f / tests/unit/CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/1e9118f41bd6d31b9c77cbc1bddc14d1d13a22c5 | 16:56 |
shogun-notifier- | shogun: Refactor autogenerated unit test creation in cmake | 16:56 |
shogun-notifier- | shogun: one will have to add manually the generation command | 16:56 |
shogun-notifier- | shogun: this way we support more flexible autogeneration | 16:56 |
@wiking | let's see what travis does with this | 16:57 |
@wiking | :> | 16:57 |
-!- iglesiasg_ [~iglesias@2001:6b0:1:1041:5158:760a:da84:6c51] has quit [Quit: Ex-Chat] | 16:57 | |
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:5dd1:908e:cfcf:27e] has joined #shogun | 16:58 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 16:58 | |
-!- HeikoS [~heiko@nat-176-108.internal.eduroam.ucl.ac.uk] has joined #shogun | 16:59 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 16:59 | |
@wiking | HeikoS: yo man | 17:00 |
@wiking | i've just sent u an email | 17:00 |
@wiking | anyhow i published the branch | 17:00 |
@wiking | and here we go we have travis as well: https://travis-ci.org/shogun-toolbox/shogun/builds/10539880 | 17:00 |
shogun-notifier- | shogun: Viktor Gal :develop * 28107fb / src/interfaces/octave_static/ (2 files): https://github.com/shogun-toolbox/shogun/commit/28107fb2b867b0c1a67f46fb6e548e1a80bba581 | 17:03 |
shogun-notifier- | shogun: Fix includes in octave_static | 17:03 |
-!- sonne|osx [~sonne@82.113.99.43] has quit [Quit: sonne|osx] | 17:10 | |
pickle27 | wiking: how can I force swig to re-update when building | 17:10 |
@wiking | what do u mean by reupdate? | 17:10 |
pickle27 | wiking: Im doing work on the swig_typemaps.i file | 17:10 |
pickle27 | wiking: but when I make changes it doesn't affect the build unless I make clean | 17:11 |
@wiking | fuck | 17:11 |
@wiking | it should | 17:11 |
@wiking | ok checking | 17:11 |
@wiking | ooooohmaaaan | 17:12 |
@wiking | clang just fucking died :) | 17:12 |
@iglesiasg | RIP | 17:12 |
@wiking | yeah | 17:12 |
@wiking | never seen a compiler fail so badly | 17:12 |
pickle27 | wiking: looks like deleting the folder under build/src/interfaces/octave_modular worked | 17:12 |
* wiking wonders why... | 17:12 | |
@wiking | pickle27: yeah that shuld definitely | 17:12 |
@wiking | but still | 17:12 |
pickle27 | it re ran a bit of configure but didn't rebuild anything | 17:13 |
@wiking | 715K Aug 23 16:53 SerializationJSON_unittest.cc mmmm | 17:13 |
@wiking | maybe it's a bit too big? :< | 17:13 |
@wiking | let's see what gcc does with it :) | 17:13 |
@wiking | pickle27: i mean the thing is that | 17:14 |
@wiking | there's this in the cmake file | 17:14 |
@wiking | COMMAND "${CMAKE_COMMAND}" -E copy_if_different ${file} ${fname} | 17:14 |
@wiking | that should make sure that if you change something in interfaces/modular/*i | 17:14 |
@wiking | then it would copy the new one | 17:14 |
pickle27 | wiking: weird | 17:14 |
@wiking | pickle27: can u check if at least it's copied over? | 17:15 |
pickle27 | wiking: I haven't upstream merged in 2 days is that line new? | 17:15 |
pickle27 | sure I'll check next time I make a change | 17:15 |
@wiking | i mean from ./src/interfaces/modular/ to ./build/src/interfaces/<whichever modular u work on> | 17:15 |
pickle27 | wiking: yeah yeah | 17:15 |
@wiking | pickle27: thnx | 17:15 |
@wiking | maybe it does copy it | 17:15 |
@wiking | but that still doesn't trigger a rebuild of modshogun.i | 17:16 |
@wiking | which i guess is going to be the case | 17:16 |
@wiking | as i would need to add some type of a dependency | 17:16 |
@wiking | of that target | 17:16 |
@wiking | let me know and i'll see | 17:16 |
pickle27 | sonney2k: around? | 17:16 |
@wiking | btW: anybody knows a way to check the peak memory used by the compiler? | 17:17 |
@wiking | not it's not ps aux|grep g++ | 17:17 |
@wiking | ? :) | 17:17 |
pickle27 | wiking: it doesn't copy | 17:19 |
@wiking | pickle27: fucker | 17:20 |
@wiking | pickle27: if u run cmake? | 17:20 |
pickle27 | wiking: I just ran make | 17:20 |
@wiking | try to run cmake | 17:20 |
@wiking | just in case | 17:20 |
pickle27 | wiking: if I run cmake I'll have t rebuild all of shogun | 17:21 |
@wiking | to see if that's effecting it | 17:21 |
@wiking | do u | 17:21 |
@wiking | ? | 17:21 |
pickle27 | I think so | 17:21 |
@wiking | that'd be weird | 17:21 |
@wiking | but ok | 17:21 |
pickle27 | I'll try | 17:21 |
@wiking | i'll check now | 17:21 |
@wiking | because it's a target that should always trigger with simple make | 17:21 |
@wiking | hahahah This log is too long to be displayed. Please reduce the verbosity of your build or download the the raw log. | 17:21 |
@wiking | ookeeey | 17:21 |
pickle27 | wiking: okay re ran cmake with the same flags didn't have t rebuild also didn't copy the file | 17:22 |
@wiking | damn cmake | 17:22 |
@wiking | ok i'll get back to u | 17:22 |
@wiking | if u have time plz fire an issue for this | 17:23 |
pickle27 | wiking: I tried deleting build/src/interface/swig_typemaps.i and then it recopied but it didn't rebuild | 17:23 |
pickle27 | wiking: sure thing | 17:24 |
-!- foulwall [~user@2001:da8:215:c252:b121:e669:c986:27a2] has joined #shogun | 17:25 | |
@wiking | tnx | 17:29 |
shogun-notifier- | shogun: Viktor Gal :feature/SerialUTests * 3455976 / tests/unit/CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/34559761f6bbbe53caa86a2ab04e2752b75e2582 | 17:29 |
shogun-notifier- | shogun: Run unit testing with --only-on-failure when executing from ctest | 17:29 |
shogun-notifier- | shogun: The number of log line in travis is limited to 10000 and the | 17:29 |
shogun-notifier- | shogun: ever growing unit tests are eating up the lines, hence only print | 17:29 |
shogun-notifier- | shogun: the failed unit tests when running it from ctest | 17:29 |
-!- dkyang [3cb13ac4@gateway/web/freenode/ip.60.177.58.196] has joined #shogun | 17:46 | |
pickle27 | wiking: where is the copy if new command in the cmake file? | 17:47 |
@wiking | cmake/CommonModularInterface.cmake | 17:49 |
@HeikoS | wiking: compiling your branch | 17:49 |
@HeikoS | ... | 17:49 |
@wiking | heheh HeikoS it's highly dangerous branch ;) | 17:50 |
@HeikoS | lets see ;) | 17:51 |
@HeikoS | wiking: I love ccache | 17:51 |
@HeikoS | finally works, my local installation somehow did not | 17:51 |
@wiking | pickle27: yeah i wonder why it does not work :S | 17:52 |
pickle27 | wiking: yeah I see it now, I don't know either | 17:52 |
pickle27 | wiking: Im able to work by just deleting the folder | 17:52 |
pickle27 | wiking: I did submit an issue on github | 17:53 |
@HeikoS | wiking: where are the tests? | 17:54 |
@HeikoS | cannot see them | 17:54 |
@wiking | HeikoS: ? | 17:54 |
@wiking | pickle27: saw it thnx | 17:54 |
@wiking | HeikoS: what do u mean where r the tests? | 17:54 |
@HeikoS | which folder? | 17:54 |
@HeikoS | or did you put everything into clone tests? | 17:54 |
@HeikoS | yep you did sorry | 17:55 |
@HeikoS | maybe rename later on | 17:55 |
@HeikoS | to automagic_unit_tests | 17:55 |
@wiking | ah | 17:57 |
shogun-notifier- | shogun: Viktor Gal :develop * f0d7298 / cmake/CommonModularInterface.cmake: https://github.com/shogun-toolbox/shogun/commit/f0d72986f40861993c1dd2b3e791b43978d253e6 | 18:01 |
shogun-notifier- | shogun: Fix for updating the .i files when they are changes | 18:01 |
shogun-notifier- | shogun: issue #1454 | 18:01 |
@wiking | pickle27: done | 18:01 |
shogun-buildbot | build #1504 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1504 blamelist: Viktor Gal <viktor.gal@maeth.com> | 18:02 |
pickle27 | wiking: cool what was the problem? | 18:02 |
@wiking | pickle27: a depend was missing ;) | 18:02 |
@wiking | rookie error | 18:02 |
@wiking | sorry about that | 18:02 |
pickle27 | wiking: np, you fixed that super fast! | 18:02 |
-!- foulwall [~user@2001:da8:215:c252:b121:e669:c986:27a2] has quit [Ping timeout: 264 seconds] | 18:06 | |
-!- travis-ci [~travis-ci@ec2-50-19-189-171.compute-1.amazonaws.com] has joined #shogun | 18:07 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/10539880 | 18:07 |
-!- travis-ci [~travis-ci@ec2-50-19-189-171.compute-1.amazonaws.com] has left #shogun [] | 18:07 | |
thoralf | wiking: "make -j$HIGH shogun shogun-unit-test install" <-- prints "Linking CXX shared library libshogun.so" 3 times and then dies with ld error "File truncated" | 18:11 |
thoralf | wiking: Can you imagine why? Are the dependencies properly set up? | 18:12 |
@wiking | thoralf: i hope so :) | 18:12 |
@wiking | thoralf: but then again ... it seems not | 18:12 |
@wiking | thoralf: what's you cmake version? | 18:12 |
thoralf | wiking: cmake version 2.8.10.1 | 18:12 |
@wiking | mmmm | 18:12 |
@wiking | i have nooo idea man | 18:12 |
thoralf | default Ubuntu 13.04 | 18:12 |
@wiking | either i was a rookie with setting up the dependencies (most probably) | 18:12 |
@wiking | or something that cmake fucks up | 18:13 |
@HeikoS | wiking: these tests are awesome, they detect everything and it always told me whats wrong | 18:14 |
@wiking | :) | 18:16 |
@wiking | so now u know why the equals fails? :) | 18:16 |
@HeikoS | yes | 18:17 |
@HeikoS | its all uninitialised memory | 18:17 |
@wiking | yey | 18:17 |
@HeikoS | people dont initialise class members | 18:17 |
@wiking | :) | 18:17 |
@wiking | fuckers :D | 18:17 |
@HeikoS | they if I ever get them | 18:19 |
@HeikoS | (so in fact just found one of my classes :D, also had some of yours a while ago ;) | 18:20 |
shogun-buildbot | build #1294 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1294 blamelist: Viktor Gal <viktor.gal@maeth.com> | 18:25 |
shogun-buildbot | build #1474 of deb2 - static_interfaces is complete: Failure [failed compile octave_static] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/1474 blamelist: Viktor Gal <viktor.gal@maeth.com> | 18:29 |
-!- dkyang [3cb13ac4@gateway/web/freenode/ip.60.177.58.196] has quit [Quit: Page closed] | 18:30 | |
@HeikoS | one more | 18:35 |
@HeikoS | wiking: around? | 18:41 |
@HeikoS | wiking: I think for those automagic tests we should activate valgrind and give errors if something is going on | 18:42 |
@HeikoS | many of those errors might go through unnoticed otherwise | 18:42 |
@wiking | HeikoS: ok | 18:42 |
@HeikoS | we are only observing the tip of the iceberg there | 18:42 |
@wiking | HeikoS: have u seen this: https://travis-ci.org/shogun-toolbox/shogun/jobs/10541373#L3717 | 18:43 |
@wiking | so there even clone_equals_MAPInference clone_equals_DisjointSet is dying as well | 18:43 |
@HeikoS | yes thats what I mean | 18:43 |
@HeikoS | its that sometimes memory errors give segfaults | 18:44 |
@HeikoS | sometimes not | 18:44 |
@HeikoS | so rather let valgrind chcek | 18:44 |
@wiking | and JSON serialization is just a big mess | 18:44 |
@wiking | :) | 18:44 |
@wiking | so we are not merging this anytime soon i have a feeling :) | 18:44 |
@HeikoS | wiking: well we can go stepwise | 18:45 |
@HeikoS | first valgrind the basic tests | 18:45 |
@HeikoS | then merge that | 18:45 |
@HeikoS | then serialization | 18:45 |
@HeikoS | then complex data | 18:45 |
@HeikoS | this is tidious fixing uninit memory bugs | 18:46 |
@wiking | HeikoS: i wanted to enable valgrind unit testing the problem was with that that openmp-ed tapkee went into infinite loop :) | 18:47 |
@HeikoS | wiking: how to print long float aka floatmaxc? | 18:47 |
@HeikoS | %f is not good? | 18:47 |
@HeikoS | wiking: maybe only the automagic ones for now? | 18:47 |
thoralf | HeikoS: %llllllf ;) | 18:47 |
@wiking | %lf | 18:47 |
@HeikoS | wiking: ah, thats a bug in clone :) | 18:47 |
@wiking | why? | 18:48 |
@HeikoS | and long int? | 18:48 |
@wiking | $ld | 18:48 |
@HeikoS | because I print the numbers sometimes when debugging | 18:48 |
@wiking | %ld | 18:48 |
@wiking | or %lld | 18:48 |
@wiking | if it's long long | 18:48 |
@HeikoS | so int64 is ld? | 18:48 |
-!- travis-ci [~travis-ci@ec2-184-72-133-65.compute-1.amazonaws.com] has joined #shogun | 18:49 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/10540255 | 18:49 |
-!- travis-ci [~travis-ci@ec2-184-72-133-65.compute-1.amazonaws.com] has left #shogun [] | 18:49 | |
@wiking | HeikoS: best is to use PRId64 | 18:49 |
@wiking | printf("%" PRId64 "\n", t); | 18:49 |
@wiking | see this | 18:49 |
@wiking | http://stackoverflow.com/questions/9225567/how-to-print-a-int64-t-type-in-c | 18:49 |
thoralf | wiking: I tried to fix the simultaneous build of libshogun by adding (random?;)) dependencies to CMakeLists, but it does not help out. | 18:52 |
@HeikoS | %Lf | 18:52 |
thoralf | wiking: Just better not use more than one target ;) | 18:52 |
@HeikoS | weiiiiird | 18:52 |
@wiking | thoralf: hehe | 18:53 |
@wiking | HeikoS: best to use those macros | 18:53 |
@HeikoS | wiking: yes, maybe I will change later | 18:53 |
@HeikoS | Parameter.cpp:35237 | 18:53 |
thoralf | HeikoS: "Maybe I will write tests later." ;) | 18:53 |
@HeikoS | thoralf: haha :) | 18:53 |
thoralf | HeikoS: Everyone knows it will be never. ;) | 18:53 |
@HeikoS | thoralf: well fixing other things | 18:53 |
@HeikoS | one after another | 18:54 |
@HeikoS | wiking: feel free to change, see file above ;) | 18:54 |
@wiking | HeikoS: okey | 18:54 |
thoralf | HeikoS: Gotta go, but if you've got time on tuesday, we could start a small hackathon and fix "all" the memory issues. | 18:54 |
@HeikoS | wiking: ok I fixed all of the problems pasted by you, sending PR | 18:55 |
@HeikoS | thoralf: brilliant :) | 18:55 |
@HeikoS | thoralf: I would agree | 18:55 |
thoralf | HeikoS: You'll be there on tuesday? | 18:55 |
@HeikoS | thoralf: dont know yet, maybe propose via email to the mailing list :) | 18:55 |
@HeikoS | then others might join too | 18:55 |
@wiking | HeikoS: cool! | 18:55 |
@wiking | HeikoS: i'm just trying to reproduce Kyle's error | 18:55 |
thoralf | Bye. | 18:56 |
-!- thoralf [~thoralf@enki.zib.de] has quit [Quit: Konversation terminated!] | 18:56 | |
@HeikoS | wiking: no the question is how to push this :) | 18:56 |
@HeikoS | getting weird git errors | 18:56 |
@wiking | HeikoS: eh? | 18:56 |
@wiking | did u rebase? | 18:56 |
@HeikoS | wiking: I have a talent to fuck up things in git | 18:56 |
@HeikoS | no i created a new branch | 18:56 |
@wiking | HeikoS: no worries... what new branch? | 18:56 |
@HeikoS | upstream/feature/SerialUTests | 18:56 |
@wiking | yeah yea | 18:56 |
@wiking | but u need to rebase | 18:57 |
@HeikoS | in there I committed | 18:57 |
@HeikoS | wiking: against what? | 18:57 |
@wiking | upstream/feature/SerialUTests | 18:57 |
@wiking | :) | 18:57 |
@HeikoS | ok let me do that | 18:57 |
@wiking | as i've pushed some stuff there in the meanwhile | 18:57 |
@HeikoS | done | 18:57 |
@HeikoS | what? | 18:57 |
@wiking | can u push now? | 18:57 |
@HeikoS | it told me up to date | 18:57 |
@HeikoS | git pull --rebase upstream feature/SerialUTests | 18:58 |
@wiking | what's the error when u try to push | 18:58 |
@HeikoS | warning: push.default is unset; its implicit value is changing in | 18:58 |
@HeikoS | Git 2.0 from 'matching' to 'simple'. To squelch this message | 18:58 |
@HeikoS | and maintain the current behavior after the default changes, use: | 18:58 |
@HeikoS | git config --global push.default matching | 18:58 |
@HeikoS | To squelch this message and adopt the new behavior now, use: | 18:58 |
@HeikoS | git config --global push.default simple | 18:58 |
@HeikoS | See 'git help config' and search for 'push.default' for further information. | 18:58 |
@HeikoS | (the 'simple' mode was introduced in Git 1.7.11. Use the similar mode | 18:58 |
@HeikoS | 'current' instead of 'simple' if you sometimes use older versions of Git) | 18:58 |
@wiking | ah ok | 18:58 |
@wiking | just do this for the moment | 18:58 |
@HeikoS | ah still worked | 18:58 |
@wiking | git push upstream feature/SerialUTests | 18:59 |
@wiking | HeikoS: i suppose that pushed to your own repo | 18:59 |
@wiking | and not the upstream ;) | 18:59 |
@HeikoS | heiko@heiko-ThinkPad-T420:~/Desktop/shogun/shogun/build$ git push upstream feature/SerialUTests | 18:59 |
@HeikoS | error: src refspec feature/SerialUTests does not match any. | 18:59 |
@HeikoS | error: failed to push some refs to 'https://github.com/shogun-toolbox/shogun.git' | 18:59 |
@wiking | eh | 18:59 |
@HeikoS | ill try a PR | 19:00 |
@HeikoS | grrr | 19:00 |
@HeikoS | I dont see the branch in github | 19:00 |
@HeikoS | at my fork | 19:00 |
@wiking | yeah | 19:00 |
@wiking | me neither | 19:01 |
@wiking | i wonder where u pushed this | 19:01 |
@HeikoS | so how to push directly to the feature branch? | 19:01 |
@HeikoS | wiking: getting this error above | 19:01 |
@wiking | mm | 19:01 |
@wiking | wait | 19:01 |
@wiking | mmm seems like that | 19:02 |
@wiking | so anyways git push upstream feature/SerialUTests | 19:02 |
@wiking | should do it | 19:02 |
@HeikoS | as said | 19:02 |
@HeikoS | heiko@heiko-ThinkPad-T420:~/Desktop/shogun/shogun/build$ git push upstream feature/SerialUTests | 19:02 |
@HeikoS | error: src refspec feature/SerialUTests does not match any. | 19:02 |
@HeikoS | error: failed to push some refs to 'https://github.com/shogun-toolbox/shogun.git' | 19:02 |
@wiking | yeah that's totally weird | 19:02 |
@wiking | as it's there :))) | 19:02 |
@wiking | HeikoS: do a git fetch | 19:02 |
@wiking | even better | 19:02 |
@HeikoS | no help | 19:03 |
@HeikoS | git fetch upstream also not help | 19:03 |
@wiking | git fetch upstream feature/SerialUTests | 19:03 |
@wiking | ? | 19:03 |
@HeikoS | that did something but still same error | 19:03 |
@HeikoS | maybe via PR | 19:04 |
@wiking | but how do u want to pr | 19:04 |
@wiking | when it's nowhere | 19:04 |
@wiking | i mean your commit is not yet anywhere in github | 19:04 |
@HeikoS | wiking: how do I push the local branch to my fork? | 19:04 |
@wiking | i guess your fork is name origin | 19:05 |
@wiking | because in that case it would be | 19:05 |
@wiking | git push origin feature/SerialUTests | 19:05 |
@HeikoS | same error | 19:05 |
@wiking | mmmm woah | 19:05 |
@HeikoS | ah maybe the ubuntu upgrade fucked up something | 19:05 |
@HeikoS | how can I check the git upstream settings? | 19:05 |
@wiking | what's the output of this | 19:06 |
@wiking | git branch | 19:06 |
@wiking | HeikoS: .git/config has all the info | 19:06 |
@HeikoS | no upstream in there just my email | 19:06 |
@HeikoS | heiko@heiko-ThinkPad-T420:~/Desktop/shogun/shogun/build$ git branch | 19:06 |
@HeikoS | dev | 19:06 |
@HeikoS | develop | 19:06 |
@HeikoS | master | 19:06 |
@HeikoS | * upstream/feature/SerialUTests | 19:06 |
@wiking | oooh fuck | 19:06 |
@HeikoS | ah is the upstream too much? | 19:06 |
@wiking | yeah created actually a branch called upstream/feature/SerialUTests :D | 19:06 |
@wiking | i think | 19:06 |
@wiking | :D | 19:06 |
@HeikoS | haha | 19:06 |
@HeikoS | nice | 19:07 |
@wiking | remote "upstream"] url = git@github.com:shogun-toolbox/shogun.git fetch = +refs/heads/*:refs/remotes/upstream/* | 19:07 |
@wiking | [remote "upstream"] url = git@github.com:shogun-toolbox/shogun.git fetch = +refs/heads/*:refs/remotes/upstream/* | 19:07 |
@wiking | this should b somewher in your .git/config | 19:07 |
@wiking | HeikoS: if u do this git branch -a | 19:07 |
@wiking | do you have something like this | 19:07 |
@wiking | remotes/upstream/feature/SerialUTests | 19:08 |
@wiking | if not then add a new remote as upstream | 19:08 |
@HeikoS | so now I should just create features/bla rather than upstream/feature/vbla? | 19:08 |
@wiking | mmm wait | 19:09 |
@wiking | do u have upstream as remote branch? | 19:09 |
shogun-notifier- | shogun: Heiko Strathmann :feature/SerialUTests * b2ca266 / src/shogun/features/streaming/generators/GaussianBlobsDataGenerator.cpp: https://github.com/shogun-toolbox/shogun/commit/b2ca26607677d92d0f7ccf8a324c0b680d81768d | 19:09 |
shogun-notifier- | shogun: fixed uninitialised memory bug | 19:09 |
shogun-notifier- | shogun: Heiko Strathmann :feature/SerialUTests * 6c4cbf8 / src/shogun/classifier/ (2 files): https://github.com/shogun-toolbox/shogun/commit/6c4cbf830f918abb0b050b339a8668ac434ba25d | 19:09 |
@HeikoS | ha I pushed | 19:09 |
shogun-notifier- | shogun: fixed uninitialized memory bugs and changed SG_REF order in setter | 19:09 |
shogun-notifier- | shogun: Heiko Strathmann :feature/SerialUTests * 58ff468 / src/shogun/multiclass/MCLDA.cpp: https://github.com/shogun-toolbox/shogun/commit/58ff468a24e44ec520e1f5a233574866be68bdda | 19:09 |
shogun-notifier- | shogun: fix uninitialised memory bugs | 19:09 |
shogun-notifier- | shogun: Heiko Strathmann :feature/SerialUTests * 8e71e1d / src/shogun/structure/DualLibQPBMSOSVM.cpp: https://github.com/shogun-toolbox/shogun/commit/8e71e1dfec58788329fcba977ae0c5aba7522889 | 19:09 |
shogun-notifier- | shogun: fix uninitialised memory bugs | 19:09 |
shogun-notifier- | shogun: Heiko Strathmann :feature/SerialUTests * 95c1414 / src/shogun/structure/Factor.cpp,src/shogun/structure/Factor.h: https://github.com/shogun-toolbox/shogun/commit/95c1414b728b01830d3cff8ca5415cf3bbea344b | 19:10 |
shogun-notifier- | shogun: fixed uninitialised memory bugs and removed unused class member factor_id (causing errors) | 19:10 |
shogun-notifier- | shogun: Heiko Strathmann :feature/SerialUTests * c784964 / src/shogun/structure/MAPInference.cpp: https://github.com/shogun-toolbox/shogun/commit/c784964444b2b046da9bc359cd953f5753407906 | 19:10 |
shogun-notifier- | shogun: fixed uninitialized memory bug | 19:10 |
shogun-notifier- | shogun: Heiko Strathmann :feature/SerialUTests * 6a6cc00 / src/shogun/base/Parameter.cpp: https://github.com/shogun-toolbox/shogun/commit/6a6cc0006ad595a2c6eac0c6e1fc378f3b41dbc2 | 19:10 |
shogun-notifier- | shogun: fixed memory bug in debug prints of clone | 19:10 |
shogun-notifier- | shogun: Heiko Strathmann :feature/SerialUTests * 1ae7e02 / src/shogun/features/StringFeatures.cpp: https://github.com/shogun-toolbox/shogun/commit/1ae7e02175cf572793fff6f7beca04fa6a579b9b | 19:10 |
shogun-notifier- | shogun: fixed uninitialized memory bugs | 19:10 |
@wiking | yey worked | 19:10 |
@wiking | ;d | 19:10 |
@wiking | :D | 19:10 |
@HeikoS | correct branch? :) | 19:10 |
@wiking | seems like | 19:10 |
@HeikoS | ok phew | 19:10 |
@HeikoS | now what next? | 19:10 |
@HeikoS | valgrind for the autotests? | 19:10 |
@HeikoS | is that easy to add? | 19:10 |
@wiking | mmm welll | 19:11 |
@wiking | oh fuck | 19:11 |
@wiking | now i have a rpoblem with my branch | 19:12 |
@HeikoS | ah git, I just love it ;) | 19:12 |
@HeikoS | eh wait | 19:13 |
@wiking | what? | 19:13 |
@HeikoS | did all my commits go in? | 19:14 |
@HeikoS | ah yes since I pushed directly | 19:14 |
@HeikoS | its not grouped | 19:14 |
@HeikoS | phew | 19:14 |
shogun-buildbot | build #1475 of deb2 - static_interfaces is complete: Failure [failed compile octave_static] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/1475 blamelist: Viktor Gal <viktor.gal@maeth.com> | 19:17 |
shogun-buildbot | build #1505 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1505 blamelist: Viktor Gal <viktor.gal@maeth.com> | 19:20 |
@HeikoS | wiking: want to see more weird stuff? | 19:22 |
@HeikoS | [ RUN ] SerializationXML.SpectrumRBFKernel | 19:22 |
@HeikoS | /tmp/SpectrumRBFKernel.PbZLhp:2: parser error : Specification mandate value for attribute sequences | 19:22 |
@HeikoS | ="float64">0</width><nof_sequences type="int32">0</nof_sequences><the sequences | 19:22 |
@HeikoS | ^ | 19:22 |
@HeikoS | /tmp/SpectrumRBFKernel.PbZLhp:2: parser error : attributes construct error | 19:22 |
@HeikoS | ="float64">0</width><nof_sequences type="int32">0</nof_sequences><the sequences | 19:22 |
@HeikoS | ^ | 19:22 |
@HeikoS | /tmp/SpectrumRBFKernel.PbZLhp:2: parser error : Couldn't find end of Start Tag the line 2 | 19:22 |
@HeikoS | ="float64">0</width><nof_sequences type="int32">0</nof_sequences><the sequences | 19:22 |
@HeikoS | ^ | 19:22 |
@HeikoS | Segmentation fault (core dumped) | 19:22 |
@wiking | HeikoS: yes | 19:23 |
@HeikoS | ah sorry | 19:23 |
@HeikoS | thats serialization | 19:23 |
@wiking | HeikoS: xml generator is not the best | 19:23 |
@wiking | i mean it's a bug in xmlwriter | 19:23 |
@wiking | i mean our XML file writer | 19:23 |
@wiking | one cannot have <fuck off> </fuck off> tag | 19:23 |
@wiking | ;) | 19:23 |
@HeikoS | wiking: i dont like the serialization tests inside the other ones | 19:24 |
@HeikoS | this segfaults | 19:24 |
@HeikoS | but I have to way to check whether the other ones worked now | 19:24 |
@HeikoS | they all should be sep. files | 19:24 |
@HeikoS | sorry to annoy you with that again | 19:24 |
@wiking | what do you mean by sep files? | 19:25 |
@wiking | --gtest_filter=SerializationHDF5\* | 19:25 |
@wiking | SystemError: [ERROR] In file /Users/wiking/shogun/src/shogun/preprocessor/PruneVarSubMean.cpp line 136: assertion initialized failed in virtual SGMatrix<float64_t> shogun::CPruneVarSubMean::apply_to_feature_matrix(shogun::CFeatures *) file /Users/wiking/shogun/src/shogun/preprocessor/PruneVarSubMean.cpp line 136 | 19:26 |
@wiking | yeye | 19:26 |
@HeikoS | I think I will go home soon | 19:27 |
@HeikoS | good efforts for today though | 19:27 |
@wiking | HeikoS: woah man | 19:27 |
@HeikoS | ? | 19:27 |
@wiking | PreprocHandle = SerializableAsciiFile('./PreprocFile.preproc', 'w') | 19:27 |
@wiking | guess what's the output of this | 19:28 |
@HeikoS | 42? | 19:28 |
@wiking | <<_SHOGUN_SERIALIZABLE_ASCII_FILE_V_00_>> | 19:28 |
@wiking | version_parameter int32 1 | 19:28 |
@wiking | this is all | 19:28 |
@HeikoS | not good | 19:28 |
@HeikoS | idea why? | 19:28 |
@wiking | heheh | 19:28 |
@wiking | there's not a single | 19:29 |
@wiking | SG_ADD | 19:29 |
@wiking | :) | 19:29 |
@wiking | so none of the parameters of that class has been added to the parameters list | 19:29 |
@wiking | :) | 19:29 |
@HeikoS | ah | 19:30 |
@HeikoS | well, so thats fine then right? :) | 19:30 |
@wiking | heheh yeah | 19:30 |
@wiking | fixing it | 19:31 |
@HeikoS | cool | 19:31 |
@HeikoS | thats kyle's error? | 19:31 |
@wiking | mmm | 19:31 |
@wiking | HeikoS: yes | 19:31 |
@HeikoS | cool then | 19:31 |
@wiking | float64_t* std; | 19:31 |
@HeikoS | ah | 19:31 |
@wiking | do we serialize this? | 19:31 |
@HeikoS | how is it used? | 19:31 |
@wiking | or rather replace this with SGVector? | 19:31 |
@HeikoS | ah yes | 19:31 |
@HeikoS | but careful about the len variables when doing this | 19:31 |
@HeikoS | and the access | 19:31 |
@wiking | okey | 19:32 |
@wiking | i'll just rewrite this class a bit | 19:32 |
@wiking | with using SGVector shit | 19:32 |
@wiking | HeikoS: SG_ADD(SGVector... works right? | 19:32 |
@HeikoS | yes but you have to give it the address | 19:32 |
@HeikoS | better check exisitng code :D | 19:32 |
@wiking | okok will do | 19:33 |
@wiking | hehe | 19:34 |
@wiking | this code is like from 20008 | 19:34 |
@wiking | :) | 19:34 |
-!- travis-ci [~travis-ci@ec2-50-19-189-171.compute-1.amazonaws.com] has joined #shogun | 19:38 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/10541372 | 19:38 |
-!- travis-ci [~travis-ci@ec2-50-19-189-171.compute-1.amazonaws.com] has left #shogun [] | 19:38 | |
@HeikoS | wiking: I edited older this year ;) | 19:39 |
@HeikoS | HMM contained uninit memory haha | 19:39 |
@HeikoS | 1999-2009 | 19:39 |
@wiking | :D | 19:42 |
shogun-buildbot | build #1295 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1295 blamelist: Viktor Gal <viktor.gal@maeth.com> | 19:44 |
-!- sonne|osx [~sonne@f053046196.adsl.alicedsl.de] has joined #shogun | 19:48 | |
pickle27 | sonne|osx: hey can I bug you about the octave swig interface? | 19:51 |
@wiking | HeikoS: how do u i SG_ADD a float64* | 19:52 |
sonne|osx | pickle27: well don't ask to ask ASK! | 19:52 |
pickle27 | haha okay | 19:52 |
pickle27 | sonne|osx: here is my NDArray type map for octave | 19:54 |
pickle27 | the correct array prints from the printf but I get a corrupted double link list error and octave raches | 19:54 |
sonne|osx | pickle27: where and which direction - in our out type map? | 19:57 |
pickle27 | sonne|osx: this is the out type map | 19:57 |
@wiking | sonne|osx: CPruneVarSubMean(bool divide=true); is actualy the implementation of CPruneVarSubMean(); ? | 19:57 |
@wiking | sonne|osx: i know it's c++ question but i'm a bit wondering now... | 19:57 |
pickle27 | sonne|osx: woops here is the paste http://pastebin.com/hr05cVqy | 19:58 |
@HeikoS | wiking: | 19:59 |
@HeikoS | what do you mean? | 19:59 |
@HeikoS | if you dont migrate it to SGVector? | 19:59 |
shogun-buildbot | build #33 of osx1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/osx1%20-%20libshogun/builds/33 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Thoralf Klein <thoralf.klein@zib.de>, Viktor Gal <viktor.gal@maeth.com> | 19:59 |
@wiking | HeikoS: yes | 19:59 |
@HeikoS | wiking: then there is sg_add_vector or so which you give the pointer address and the length pointer | 19:59 |
@wiking | HeikoS: SG_ADD(m_mean, "mean_vec", "Mean vector", MS_NOT_AVAILABLE); | 20:00 |
@wiking | not good? | 20:00 |
@wiking | where float64_t* m_mean | 20:00 |
@HeikoS | no | 20:00 |
@HeikoS | bad | 20:00 |
@wiking | fucker | 20:00 |
@wiking | there's no SG_ADD_V | 20:00 |
@HeikoS | sg_add_vector(&m_mean, | 20:00 |
@wiking | HeikoS: no such thing | 20:01 |
@wiking | git grep sg_add_v | 20:01 |
@HeikoS | let me find the right one | 20:01 |
@HeikoS | Parameter::add_vector( | 20:01 |
@wiking | oooh yeah segfault | 20:01 |
pickle27 | sonne|osx: I also don't really know how to debug this, I tried running valgrind on octave here is the output | 20:02 |
@HeikoS | ? | 20:02 |
@HeikoS | wiking: registering old style vectors is tricky | 20:02 |
@HeikoS | best check existing code | 20:02 |
@HeikoS | wiking: better: migrate to SGVector | 20:02 |
-!- lisitsyn [~lisitsyn@109-226-90-47.clients.tlt.100megabit.ru] has joined #shogun | 20:03 | |
@HeikoS | wiking: better better: do this and add unit test :D | 20:03 |
sonne|osx | pickle27: where is here? | 20:03 |
pickle27 | sonne|osx: here is the paste: http://pastebin.com/hr05cVqy | 20:04 |
@wiking | HeikoS: m_idx = NULL; what would be the equivalent for SGVector>? | 20:04 |
@wiking | HeikoS: resize_vector(0) ? | 20:04 |
lisitsyn | pickle27: oh I am just in time | 20:04 |
@wiking | :) | 20:04 |
@HeikoS | no | 20:04 |
@HeikoS | wiking: what is idx? | 20:04 |
@wiking | HeikoS: then? | 20:04 |
pickle27 | sonne|osx: oh the valgrind, its coming in a second | 20:05 |
@wiking | SGVector<int32_t> | 20:05 |
@HeikoS | you can set the vector to empty with m_idx=SGVector<...>(0) | 20:05 |
@HeikoS | or better () | 20:05 |
pickle27 | lisitsyn: Im working on added NDArray to the octave modular interface | 20:05 |
pickle27 | adding* | 20:05 |
lisitsyn | pickle27: what about python? | 20:05 |
pickle27 | lisitsyn: python is done | 20:05 |
lisitsyn | cool | 20:06 |
@wiking | HeikoS: ? | 20:06 |
@sonney2k | pickle27, looks good to me | 20:06 |
@HeikoS | m_idx=SGVector<bla>() | 20:06 |
@wiking | HeikoS: m_mean = SGVector<float64_t>(); | 20:06 |
@HeikoS | unrefs existing, sets array to NULL and length to 0 | 20:06 |
@sonney2k | pickle27, I mean the valgrind output - where is it? | 20:06 |
@HeikoS | yes | 20:06 |
@HeikoS | this should be equivalent to array=NULL, array_len=0 | 20:07 |
pickle27 | sonney2k: I am just rebuilding so I can run it, I had some other debug stuff in there | 20:07 |
@HeikoS | in old style | 20:07 |
@wiking | HeikoS: would that be calling free on the inner struct? | 20:09 |
@HeikoS | wiking: yes since it unrefs it | 20:09 |
pickle27 | sonney2k: here is the valgrind http://pastebin.com/4PW6jJDg | 20:09 |
@HeikoS | but only if you previously used SGVector to handle the structure | 20:09 |
@HeikoS | =no hacks turning arrays into SGVectors | 20:10 |
pickle27 | sonney2k: and I get this error if running in octave normaly | 20:11 |
pickle27 | *** Error in `octave': corrupted double-linked list: 0x00000000018060e0 *** | 20:11 |
@sonney2k | pickle27, why do you do dim_vector vdims(1, num_dims); ? | 20:11 |
-!- lambday [67157c37@gateway/web/freenode/ip.103.21.124.55] has joined #shogun | 20:11 | |
@sonney2k | pickle27, shouldn't it be dim_vector vdims(num_dims); ? | 20:11 |
lambday | HeikoS: hi | 20:11 |
@HeikoS | lambday: hey! | 20:12 |
@HeikoS | how is it going? | 20:12 |
lambday | HeikoS: phew! just got out of the lab! | 20:12 |
@HeikoS | lambday: just ported the ozone sampler to python | 20:12 |
lambday | HeikoS: probing is almost done | 20:12 |
@HeikoS | lambday: and I got the PBS computation engine working in python | 20:12 |
@HeikoS | lambday: wow! | 20:12 |
@HeikoS | cool! | 20:12 |
lambday | HeikoS: do you think we should keep the matrix power module in another class hierarchy ? | 20:12 |
pickle27 | sonney2k: its gives a deprecated warning if I do that, plus it was done with dim_vector(1,len) for the SGVector typemap | 20:12 |
@HeikoS | what are you working on? | 20:12 |
@HeikoS | lambday: yes, in CMath | 20:12 |
lambday | HeikoS: as a static method? | 20:13 |
@HeikoS | lambday: I guess thats best? | 20:13 |
lambday | HeikoS: wow!!!! | 20:13 |
pickle27 | sonney2k: so I just did the same waya | 20:13 |
lambday | HeikoS: yes.. I think so.. | 20:13 |
@sonney2k | pickle27, doesn't make sense to me | 20:13 |
lambday | I think I can manage it with log(n) multiplication instead of n for power n of a matrix | 20:13 |
pickle27 | sonney2k: yeah confuses me too, but both Vector and Matrix typemaps are like that | 20:13 |
lambday | at the cost of O(1) extra storage | 20:14 |
pickle27 | also I can't even find the docs for the dim_vector(int, int) constructor its not in doxygen | 20:14 |
pickle27 | for octave | 20:14 |
lambday | but since its just once every time LogDetEstimator is run (and only within the precompute of the sampler), we can afford that | 20:14 |
@HeikoS | lambday: sweet! | 20:14 |
@sonney2k | pickle27, you should really use dim_vector(dims) | 20:14 |
@HeikoS | lambday: thats really good news | 20:15 |
@iglesiasg | see you later people | 20:15 |
lambday | HeikoS: so, regarding multiple implementation of matrix power, how should I do that?? | 20:15 |
pickle27 | sonney2k: do you mean dims or num_dims? | 20:15 |
@HeikoS | lambday: Daniel and the others never solved it that nice :) | 20:15 |
@HeikoS | iglesiasg: bye! | 20:15 |
@sonney2k | pickle27, maybe dim_vector.alloc(dims) is the new way but certainly not dim_vector(1,foo) or so | 20:15 |
@HeikoS | lambday: what do you mean? | 20:15 |
@sonney2k | pickle27, num_dims | 20:15 |
@sonney2k | pickle27, try dim_vector.alloc(num_dism) | 20:15 |
pickle27 | sonney2k: okay | 20:15 |
lambday | HeikoS: I am not too sure :( let me finish it first then I'll celebrate with Rum and Beer :D | 20:15 |
@HeikoS | haha :D | 20:16 |
lambday | HeikoS: we will have two different implementations of matrix powers, right? | 20:16 |
@HeikoS | lambday: will be going for beer soon, too | 20:16 |
@HeikoS | lambday: which are? | 20:16 |
lambday | lol and I am already high in rum | 20:16 |
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:5dd1:908e:cfcf:27e] has quit [Quit: Ex-Chat] | 20:16 | |
lambday | HeikoS: direct one and the eigenvalue one | 20:16 |
@HeikoS | lambday: ah | 20:16 |
@HeikoS | ok | 20:16 |
@HeikoS | lambday: direct one is just std product right? | 20:16 |
@sonney2k | lambday, :: of course instead of . | 20:16 |
lambday | sonney2k: HeikoS.. no no :-/ | 20:17 |
lambday | I mean, I can't think of two different signatures | 20:17 |
@HeikoS | lambday: I am confused | 20:17 |
lambday | except for a param | 20:17 |
@sonney2k | lambday, what? | 20:17 |
@HeikoS | lambday: make it two methods | 20:18 |
lambday | errr I am confused two | 20:18 |
lambday | yes | 20:18 |
lambday | and then an extra param | 20:18 |
lambday | for one | 20:18 |
lambday | right? | 20:18 |
@HeikoS | ye | 20:18 |
@HeikoS | p | 20:18 |
@HeikoS | ah | 20:18 |
@HeikoS | I see but you want it to be polymorphic | 20:18 |
lambday | I don't like it - that is the problem | 20:18 |
lambday | yes :D | 20:18 |
@HeikoS | common interface | 20:18 |
@HeikoS | lambday: then make it a class | 20:18 |
@HeikoS | in math | 20:18 |
@sonney2k | pickle27, I am pretty confident that this will work | 20:18 |
lambday | in math itself? | 20:18 |
@HeikoS | thats nicer anyway | 20:18 |
@HeikoS | no | 20:18 |
@HeikoS | in the folder | 20:19 |
lambday | alright yes I was thinking about that | 20:19 |
@wiking | HeikoS: vector = new_vector | 20:19 |
@sonney2k | pickle27, ohh and please change the other dim_vectors too then to fix the deprecated... | 20:19 |
@wiking | HeikoS: thi sis releasing the old memory + and refs new_vector? | 20:19 |
lambday | I thought of fitting in current COperatorFunction only but it won't fit | 20:19 |
@HeikoS | wiking: yes | 20:19 |
@HeikoS | wiking: its all copy by value but still magic | 20:19 |
pickle27 | sonney2k: if it works for sure | 20:19 |
@sonney2k | pickle27, so does it work? | 20:19 |
pickle27 | sonney2k: still building | 20:20 |
pickle27 | sonney2k: rebuilding modular takes a while | 20:20 |
@sonney2k | it all takes too long now | 20:20 |
@sonney2k | it the more c++y we write the code the longer it will take | 20:20 |
pickle27 | sonney2k: looks like the same bug as before | 20:21 |
@HeikoS | lambday: maybe thats good | 20:21 |
@HeikoS | sparsematrix operator? | 20:21 |
lambday | HeikoS: today I had a meeting with Intel US people, and surprisingly my supervisor introduced me as the GSoC guy (in a sense that its a strong point to support that I can do a little bit of programming) | 20:21 |
lambday | lol... :D | 20:21 |
lambday | surprising :D | 20:21 |
@HeikoS | lambday: cool! | 20:21 |
@HeikoS | lambday: so they like it after all :) | 20:21 |
@sonney2k | pickle27, cannot be? you sure it took the right lib? | 20:22 |
@HeikoS | you should show them your awesome framework to support that | 20:22 |
@sonney2k | pickle27, .so I mean? | 20:22 |
lambday | HeikoS: let's hope so.. :D | 20:22 |
pickle27 | sonney2k: I rebuilt and ran sudo make install again | 20:22 |
lambday | HeikoS: so different class it is... | 20:22 |
lambday | alright | 20:23 |
@HeikoS | lambday: no why not linear operator? | 20:23 |
@sonney2k | pickle27, then please show the valgrind output | 20:23 |
@HeikoS | just has to be documented | 20:23 |
lambday | within linear operator itself? | 20:23 |
lambday | umm yeah that might be a good idea! | 20:23 |
lambday | and also, the CSparseMatrixOperator will have a method that returns its sparsity structure that will be passed to the coloring interface | 20:24 |
@HeikoS | lambday: yeah why not, might be useful | 20:24 |
lambday | sexy! | 20:24 |
pickle27 | sonney2k: http://pastebin.com/wqge0B6q | 20:24 |
@HeikoS | lambday: cool! | 20:24 |
pickle27 | sonney2k: I am going to try with a int_32 cast on the vdim loop | 20:25 |
@sonney2k | pickle27, try to use num_dims+1 | 20:25 |
@sonney2k | when allocating the vector | 20:26 |
@sonney2k | pickle27, the cast won't help | 20:26 |
lambday | HeikoS: hopefully these all be done by this week.. | 20:26 |
@sonney2k | and doesn't hurt either though | 20:26 |
pickle27 | okay | 20:26 |
lambday | HeikoS: oh and wiking and I were talking about the parallel implementation of our computation engine... | 20:26 |
pickle27 | I'll try the cast and then try the +1 | 20:26 |
@HeikoS | lambday: yes a bit | 20:26 |
@sonney2k | lambday, btw please make n a int64_t | 20:26 |
@HeikoS | lambday: as said, I am writing the thing in python to experiment with it | 20:26 |
@HeikoS | will share soon, once it is working | 20:26 |
lambday | sonney2k: n as in matrix power? | 20:27 |
@HeikoS | its more or less your interface | 20:27 |
@sonney2k | matrices might be bigger but on the other hand IDK whether octave supports 64bit matrices | 20:27 |
@sonney2k | lambday, prod(dims(tensor)) == num_elements | 20:27 |
@sonney2k | no :) | 20:27 |
lambday | tensor?? :-o | 20:27 |
* lambday drops his jaws | 20:28 | |
@sonney2k | n-dimensional array if you want to relax :) | 20:28 |
lambday | HeikoS: wiking was suggesting that hadoop can be a better alternative than mpi | 20:28 |
@HeikoS | lambday: yep might be | 20:28 |
@HeikoS | lambday: in fact, I also just used FS rather than sockets etc | 20:28 |
@HeikoS | lets start with PBS | 20:28 |
@wiking | lambday: not better just another :) | 20:28 |
@HeikoS | since many people have access to that | 20:29 |
@HeikoS | most useful | 20:29 |
@HeikoS | although a bit of overhead in there | 20:29 |
lambday | sonney2k: I am confused :-/ you meant the matrix datatype to be of int64 or the matrix dim to be int64? :( | 20:29 |
lambday | HeikoS: I don't have much idea regarding this... I am hoping I'll learn as we start implementing it | 20:30 |
@HeikoS | lambday: you will, dont worry | 20:30 |
shogun-notifier- | shogun: Viktor Gal :develop * 4917af6 / src/shogun/preprocessor/ (2 files): https://github.com/shogun-toolbox/shogun/commit/4917af6c0a241f1eb0b7b7945c1d2c768d9c8b3b | 20:30 |
shogun-notifier- | shogun: Refactor PruneVarSubMean | 20:30 |
shogun-notifier- | shogun: Use SGVector instead of simepl arrays and register all the parameters of | 20:30 |
shogun-notifier- | shogun: the class in order to support serialization | 20:30 |
@HeikoS | lambday: Ill share some python code soon | 20:30 |
@wiking | there | 20:30 |
lambday | HeikoS: awesome! :D | 20:31 |
sonne|osx | wiking: woah! | 20:31 |
sonne|osx | you are on a spray | 20:31 |
lambday | HeikoS: oh and roman added a get_diagonal in SGMatrix only, right? so I think I can use those in my CDenseMatrixOperator | 20:31 |
sonne|osx | with the keyboard being your weapon | 20:31 |
lambday | sonne|osx: you're on mac yay :D | 20:31 |
@HeikoS | lambday: yes | 20:31 |
@HeikoS | lambday: keep in mind they copy though | 20:31 |
@HeikoS | lambday: so not good for computing trace | 20:32 |
lambday | HeikoS: as of now I'll leave everything as it is since they work.. | 20:32 |
sonne|osx | lambday: I am on all arches now | 20:32 |
sonne|osx | archs | 20:32 |
@HeikoS | lambday: yes better | 20:33 |
lambday | later I'll think of optimization | 20:33 |
sonne|osx | 3 machines running :/ | 20:33 |
lambday | sonne|osx: you are God :) | 20:33 |
sonne|osx | which reminds me I should order that ovh machine | 20:33 |
@HeikoS | dude too many computers | 20:33 |
sonne|osx | lambday: yeah I feel like one | 20:33 |
sonne|osx | ;) | 20:33 |
sonne|osx | alright I take a short break | 20:33 |
lambday | and I can't even fuckin' buy my Thinkpad :'( | 20:33 |
sonne|osx | pickle27: post your news if you have any! | 20:34 |
@wiking | yey | 20:34 |
@wiking | HeikoS: we need such a test for serialization :) | 20:34 |
@wiking | HeikoS: i mean no sane class would have an empty file | 20:35 |
@wiking | HeikoS: although not so true | 20:35 |
@wiking | mmm dunno how we could test this :) | 20:35 |
@HeikoS | wiking: thats easy, test that nzumber of registered parameters is positive :) | 20:36 |
@wiking | sonne|osx: there's a problem with XML serialization implementation | 20:36 |
@wiking | sonne|osx: xml schema does not allow to have <tag name>.... as name would be taken as an attribute | 20:36 |
@HeikoS | wiking: but we cannot check if class members are registered | 20:36 |
@HeikoS | impossible | 20:36 |
@wiking | sonne|osx: but we are doing like <tag name></tag name> | 20:36 |
@wiking | and obviously libxml freaks out :) | 20:37 |
@wiking | hehe | 20:37 |
@wiking | https://github.com/shogun-toolbox/shogun/pulse | 20:37 |
@wiking | ;) | 20:37 |
shogun-buildbot | build #1476 of deb2 - static_interfaces is complete: Failure [failed compile octave_static] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/1476 blamelist: Viktor Gal <viktor.gal@maeth.com> | 20:39 |
shogun-buildbot | build #34 of osx1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/osx1%20-%20libshogun/builds/34 blamelist: Viktor Gal <viktor.gal@maeth.com> | 20:40 |
shogun-buildbot | build #1506 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1506 blamelist: Viktor Gal <viktor.gal@maeth.com> | 20:41 |
-!- travis-ci [~travis-ci@ec2-23-22-13-64.compute-1.amazonaws.com] has joined #shogun | 20:45 | |
travis-ci | [travis-ci] it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/10545089 | 20:45 |
-!- travis-ci [~travis-ci@ec2-23-22-13-64.compute-1.amazonaws.com] has left #shogun [] | 20:45 | |
pickle27 | sonney2k: the +1 didn't fix it | 20:50 |
pickle27 | sonney2k: I have to step out for a bit but if you have any ideas I'll check the logs | 20:50 |
@HeikoS | lambday: going home, see you later! :) | 20:55 |
@HeikoS | bye everyone! | 20:55 |
-!- HeikoS [~heiko@nat-176-108.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 20:55 | |
@wiking | pickle27: here? | 21:00 |
@wiking | pickle27: u seemed to have a deeper understanding of octave. do you have any ideas about this error: http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/1476/steps/compile%20octave_static/logs/stdio | 21:01 |
shogun-buildbot | build #1296 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1296 blamelist: Viktor Gal <viktor.gal@maeth.com> | 21:03 |
@wiking | heheheh cool | 21:04 |
@wiking | integration test failed with prunesubvar | 21:04 |
@wiking | https://travis-ci.org/shogun-toolbox/shogun/jobs/10548060 | 21:04 |
@wiking | ideas: https://travis-ci.org/shogun-toolbox/shogun/jobs/10548060#L3655 | 21:06 |
@wiking | ah ok | 21:12 |
@wiking | i get it data should be updated :) | 21:12 |
pickle27 | wiking: | 21:14 |
pickle27 | the deprecated error should be changed to dim_vector.alloc(int n) rather than using the constructor | 21:14 |
pickle27 | I don't know what the plus is about | 21:14 |
@wiking | pickle27: what's octave_allocation_error? | 21:15 |
@wiking | curr_sym_tab | 21:15 |
@wiking | global_sym_tab | 21:15 |
@wiking | are these some globals in octave? | 21:15 |
pickle27 | wiking: I really have no idea | 21:16 |
@wiking | heheh ok | 21:16 |
@wiking | i'll see | 21:16 |
@wiking | oh maaan | 21:17 |
@wiking | i need to do some regexp on that #define OCTAVE_APIVERSION api-v48+ | 21:17 |
@wiking | :P | 21:17 |
@wiking | as i have like #if OCTAVE_APIVERSION < 37 | 21:17 |
@wiking | in src/interfaces/octave_static/OctaveInterface.cpp | 21:18 |
@wiking | `octave-config -p API_VERSION | tr -d 'a-z-+'` | 21:18 |
@wiking | hehe this hack | 21:18 |
pickle27 | wiking: those errors looked more like static interface errors though | 21:20 |
pickle27 | see if any of that stuff is defined in there | 21:20 |
@wiking | pickle27: now i know why it's errored | 21:20 |
pickle27 | wiking: yah? | 21:21 |
@wiking | the way i defined the OCTAVE_APIVERSION is not the way it was in the old ./configure | 21:21 |
@wiking | i have to fix that with a regexp | 21:21 |
shogun-buildbot | build #1590 of deb3 - modular_interfaces is complete: Failure [failed test python_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1590 blamelist: Viktor Gal <viktor.gal@maeth.com> | 21:26 |
-!- travis-ci [~travis-ci@ec2-23-22-13-64.compute-1.amazonaws.com] has joined #shogun | 21:31 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/10548057 | 21:31 |
-!- travis-ci [~travis-ci@ec2-23-22-13-64.compute-1.amazonaws.com] has left #shogun [] | 21:31 | |
shogun-notifier- | shogun: Viktor Gal :develop * 2f9be65 / cmake/FindOctave.cmake: https://github.com/shogun-toolbox/shogun/commit/2f9be6517cdbf08f705f6819d7709977c280a6e7 | 21:31 |
shogun-notifier- | shogun: Fix OCTAVE_API_VERSION macro value | 21:31 |
@wiking | let's hope this one will fix it | 21:31 |
@sonney2k | wiking, any idea how I can disable variable name substitution in cmake? | 21:32 |
@wiking | sonney2k: what do u mean? :) | 21:32 |
shogun-buildbot | build #1589 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1589 blamelist: Viktor Gal <viktor.gal@maeth.com> | 21:32 |
@wiking | sonney2k: u dont want to allow that some variable can be redefined from cmd line? | 21:33 |
@sonney2k | wiking, if I have if ( var_name MATCHES foo ) then varname might expand to the variable content | 21:33 |
@sonney2k | wiking, no | 21:33 |
@wiking | sonney2k: oh... i thought only if you do ${varname} | 21:33 |
@wiking | that's only when it expands | 21:33 |
sonne|osx | wiking: no | 21:34 |
@wiking | but seems not | 21:34 |
@wiking | :( | 21:34 |
@sonney2k | wiking, http://www.cmake.org/Wiki/CMake/Language_Syntax | 21:34 |
@sonney2k | When CMake wants a variable name, when it wants a substitution. | 21:34 |
@sonney2k | wiking, hmmhh whatever maybe I don't need to solve this | 21:35 |
@wiking | what? :D | 21:35 |
@wiking | sonney2k: fix the integration test plz | 21:35 |
@sonney2k | wiking, what? | 21:35 |
@wiking | sonney2k: as i cannot generate the right data on my osx :S | 21:36 |
@sonney2k | who did break them? | 21:36 |
@wiking | sonney2k: well now that i've fixed prunesubvar | 21:36 |
@wiking | obviously the deserialization file changed | 21:36 |
@wiking | as until now it was empty basically | 21:36 |
@wiking | ;) | 21:36 |
@wiking | sonney2k: so i 'broke' them :) | 21:36 |
@wiking | but then again this was expected as prunesubvarmean was not saving anything until now ;P | 21:37 |
@wiking | s/saving/deserializing/g | 21:37 |
@wiking | sonney2k: the diff speaks for itself: https://travis-ci.org/shogun-toolbox/shogun/jobs/10548060#L3743 | 21:38 |
shogun-buildbot | build #1477 of deb2 - static_interfaces is complete: Failure [failed compile r_static] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/1477 blamelist: Viktor Gal <viktor.gal@maeth.com> | 21:39 |
@wiking | rstatic that is :) | 21:39 |
shogun-buildbot | build #35 of osx1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/osx1%20-%20libshogun/builds/35 blamelist: Viktor Gal <viktor.gal@maeth.com> | 21:40 |
shogun-buildbot | build #1507 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1507 blamelist: Viktor Gal <viktor.gal@maeth.com> | 21:40 |
@wiking | sonney2k: btw the _static interfaces shoudl be compiled with what flags? | 21:42 |
@wiking | e.g. -O3 ... etc? or -O0 | 21:42 |
shogun-notifier- | shogun: Viktor Gal :develop * ede874a / src/interfaces/r_static/ (3 files): https://github.com/shogun-toolbox/shogun/commit/ede874a2aa0a709ba2023285a038a7cd5cfbc133 | 21:42 |
shogun-notifier- | shogun: Fix r_static interface includes and compiler flags | 21:42 |
@wiking | mmm i'm commiting like a crazy monkey today | 21:42 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * e251cde / CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/e251cde8ef51616bc71bd55ba108072425ae5fcc | 21:45 |
shogun-notifier- | shogun: add summary to cmake | 21:45 |
@sonney2k | wiking, same flags like libshogun but it doesn't really make a difference | 21:46 |
@sonney2k | wiking, agreed crazy monkey | 21:46 |
pickle27 | sonney2k: I think I found the problem | 21:48 |
shogun-buildbot | build #1478 of deb2 - static_interfaces is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/1478 | 21:48 |
pickle27 | sonney2k: for some reasons the number ndims property of the octave object isn't being set properly | 21:48 |
shogun-buildbot | build #36 of osx1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/osx1%20-%20libshogun/builds/36 blamelist: Viktor Gal <viktor.gal@maeth.com> | 21:49 |
shogun-buildbot | build #1508 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1508 blamelist: Viktor Gal <viktor.gal@maeth.com> | 21:50 |
@sonney2k | wiking, so why don't you generate data then? | 21:51 |
@sonney2k | just do python generator.py | 21:51 |
@wiking | sonney2k: nogod.... | 21:53 |
@wiking | sonney2k: no good... i.e. integration test fails on osx | 21:53 |
-!- lambday [67157c37@gateway/web/freenode/ip.103.21.124.55] has quit [Ping timeout: 250 seconds] | 21:53 | |
shogun-buildbot | build #1297 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1297 blamelist: Viktor Gal <viktor.gal@maeth.com> | 22:04 |
shogun-notifier- | shogun: Viktor Gal :develop * fc2e243 / CMakeLists.txt,cmake/FindMatlab.cmake: https://github.com/shogun-toolbox/shogun/commit/fc2e24361805b875061c748224747c64ef513afa | 22:15 |
shogun-notifier- | shogun: Add target for matlab_static | 22:15 |
shogun-notifier- | shogun: add a more advanced FindMatlab.cmake script for finding Matlab libraries | 22:15 |
shogun-notifier- | shogun: and headers | 22:15 |
@wiking | and we have that one as well... | 22:15 |
@wiking | now it's 'just' about adding the tests... | 22:16 |
@sonney2k | wiking, I still have just parse error | 22:22 |
@sonney2k | wiking, what do you mean? | 22:22 |
@sonney2k | pickle27, did you get it fixed? | 22:22 |
pickle27 | sonney2k: not yet | 22:22 |
@wiking | sonney2k: ? | 22:22 |
@wiking | sonney2k: parse errors of...? | 22:23 |
@sonney2k | wiking, what is not working exactly? | 22:23 |
pickle27 | sonney2k: looking at some octave source right now because the dim_vector is not actually 3 dimensions even though it should be given the argument to alloc | 22:23 |
shogun-buildbot | build #1592 of deb3 - modular_interfaces is complete: Failure [failed test python_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1592 blamelist: Viktor Gal <viktor.gal@maeth.com> | 22:23 |
pickle27 | here is the source | 22:23 |
pickle27 | http://octave.sourceforge.net/doxygen/html/df/dd3/dim-vector_8h_source.html#l00236 | 22:23 |
pickle27 | sonney2k: am I correct in thinking this means that when n is less than 2 the vector is 2 in length | 22:24 |
@wiking | sonney2k: integration test for prunevarmean needs to be fixed.... | 22:24 |
pickle27 | sonney2k: or is it the other way and if its bigger than 2 they force it to 2 because that te behaviour Im seeing | 22:24 |
@wiking | sonney2k: and on osx the integration test are failing... regardless of what's the state now... so we need a debian/linux machine to generate those tests | 22:25 |
@sonney2k | wiking, yes so please run python generator.py name_of_prunevarsubmean.py | 22:25 |
@wiking | sonney2k: i dont have linux :( | 22:25 |
@wiking | sonney2k: if i generate it on my osx it'll fail on all the bots | 22:25 |
@sonney2k | wiking, ah so why does it fail on osx then? | 22:25 |
@wiking | sonney2k: well it's a good question but it fails all the time... i mean it always failed | 22:26 |
@wiking | sonney2k: i think it's some rounding error or something | 22:26 |
shogun-buildbot | build #37 of osx1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/osx1%20-%20libshogun/builds/37 blamelist: Soeren Sonnenburg <sonne@debian.org>, Viktor Gal <viktor.gal@maeth.com> | 22:30 |
shogun-buildbot | build #1298 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1298 blamelist: Viktor Gal <viktor.gal@maeth.com> | 22:31 |
shogun-buildbot | build #1509 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1509 blamelist: Soeren Sonnenburg <sonne@debian.org>, Viktor Gal <viktor.gal@maeth.com> | 22:31 |
@wiking | heh we have an error in structure/BeliefPropagation.cpp | 22:34 |
@wiking | oh woah since when do we allow to have #include <vector> and #include <set> in .h? | 22:38 |
@sonney2k | we don't but D-ptrs to the rescue! | 22:38 |
@wiking | sonney2k: shogun/structure/BeliefPropagation.h | 22:40 |
@wiking | it's full with STL | 22:40 |
pickle27 | sonney2k: fixed it! | 22:45 |
pickle27 | twas a silly mistake by me | 22:45 |
@sonney2k | pickle27, what was it? | 22:48 |
pickle27 | sonney2k: dim_vector.alloc returns a dim_vector so I was just using it wrong | 22:49 |
@sonney2k | pickle27, errm what did you write and you do write now? | 22:49 |
@sonney2k | code lines please | 22:49 |
pickle27 | before I had dim_vector vdim; | 22:50 |
pickle27 | vdim.alloc(num_dims) | 22:50 |
pickle27 | now I have | 22:50 |
pickle27 | dim_vector vdim = dim_vector::alloc(num_dims); | 22:50 |
@sonney2k | pickle27, ahh ok | 22:50 |
@sonney2k | sure thing then | 22:50 |
@sonney2k | pickle27, good that you realized the mistake! | 22:50 |
pickle27 | sonney2k: it works now although octave squashes the ND array back into 2D | 22:51 |
pickle27 | sonney2k: can you have 3D array's in octave? you must be able to | 22:51 |
pickle27 | sonney2k: okay yes you can have 3D arrays in octave | 22:52 |
pickle27 | sonney2k: looks like I need to do something different for this wrapper, but it does work without error now | 22:52 |
@sonney2k | pickle27, and ndims(your_array) says what? | 22:53 |
@sonney2k | pickle27, | 22:54 |
@sonney2k | a(1,1,1,1,1,1,1)=1 | 22:54 |
@sonney2k | a = 1 | 22:54 |
@sonney2k | octave:3> ndims(a) | 22:54 |
@sonney2k | ans = 2 | 22:54 |
@sonney2k | a(1,1,1,1,1,1,2)=1 | 22:54 |
@sonney2k | ndims(a) | 22:54 |
@sonney2k | ans = 7 | 22:54 |
@sonney2k | so it depends how your data looks like | 22:54 |
@sonney2k | wiking, no idea I am unable to commit | 22:55 |
@sonney2k | wiking, push I mean | 22:55 |
@wiking | sonney2k: why not/ | 22:55 |
pickle27 | sonney2k: it should be 2x2x4 but it comes out as 2x8 | 22:55 |
@sonney2k | new setup is still WIP | 22:55 |
@wiking | :D | 22:55 |
@sonney2k | pickle27, well try some 5d matrix or so with the latest dim > 1 | 22:56 |
@sonney2k | anyway bed time | 22:56 |
@sonney2k | cu | 22:56 |
pickle27 | later | 22:56 |
shogun-buildbot | build #1299 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1299 blamelist: Soeren Sonnenburg <sonne@debian.org>, Viktor Gal <viktor.gal@maeth.com> | 22:58 |
-!- travis-ci [~travis-ci@ec2-184-72-133-65.compute-1.amazonaws.com] has joined #shogun | 23:03 | |
travis-ci | [travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/10550407 | 23:03 |
-!- travis-ci [~travis-ci@ec2-184-72-133-65.compute-1.amazonaws.com] has left #shogun [] | 23:03 | |
shogun-buildbot | build #1593 of deb3 - modular_interfaces is complete: Failure [failed test python_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1593 blamelist: Soeren Sonnenburg <sonne@debian.org>, Viktor Gal <viktor.gal@maeth.com> | 23:03 |
pickle27 | sonney2k: btw got it fixed, just needed to change the macros slightly | 23:14 |
-!- travis-ci [~travis-ci@ec2-184-72-133-65.compute-1.amazonaws.com] has joined #shogun | 23:25 | |
travis-ci | [travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/10550486 | 23:25 |
-!- travis-ci [~travis-ci@ec2-184-72-133-65.compute-1.amazonaws.com] has left #shogun [] | 23:25 | |
shogun-buildbot | build #1591 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1591 blamelist: Viktor Gal <viktor.gal@maeth.com> | 23:55 |
--- Log closed Sat Aug 24 00:00:28 2013 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!