Open in new window / Try shogun cloud
--- Log opened Sun Mar 17 00:00:13 2013
-!- shogun-notifier- [] has quit [Quit: transmission timeout]00:26
-!- heiko [] has joined #shogun01:32
-!- heiko [] has left #shogun []01:50
-!- shogun-notifier- [] has joined #shogun02:18
shogun-notifier-shogun: Sergey Lisitsyn :master * ce73ba2 / src/shogun/lib/tapkee/ (19 files):
shogun-notifier-shogun: Last update of tapkee before release02:18
shogun-notifier-shogun: - Much more doc02:18
shogun-notifier-shogun: - Various code improvements (better naming, less warnings)02:18
-!- heiko [] has joined #shogun02:28
-!- heiko [] has left #shogun []02:52
shogun-buildbot_build #358 of rpm1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Sergey Lisitsyn <>03:33
-!- votjak [] has joined #shogun03:50
-!- shogun-notifier- [] has quit [Quit: transmission timeout]05:18
-!- hoijui [] has joined #shogun06:40
@sonney2kblackburn, make prepare-release ran through here!08:11
@sonney2kblackburn, so we could release - but question is if we don't better fix the label bug in static interfaces?08:11
@sonney2kalways returning real valued outputs is important08:12
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun09:00
-!- shogun-notifier- [] has joined #shogun09:47
shogun-notifier-shogun: Soeren Sonnenburg :master * 731cbde / src/ChangeLog:
shogun-notifier-shogun: updated changelog09:47
shogun-notifier-shogun: Soeren Sonnenburg :master * b97829a / doc/pages/ (4 files):
shogun-notifier-shogun: updated reference documentation09:47
shogun-notifier-shogun: Soeren Sonnenburg :master * cc2410f / src/shogun/ui/SGInterface.cpp:
shogun-notifier-shogun: support various label types in static interface09:47
-!- n4nd0 [] has joined #shogun10:02
-!- sidi_ [~sidi@] has joined #shogun10:10
-!- zxtx [] has joined #shogun10:10
shogun-buildbot_build #815 of deb2 - static_interfaces is complete: Failure [failed test octave_static]  Build details are at  blamelist: Soeren Sonnenburg <>10:12
-!- n4nd0 [] has quit [Quit: leaving]10:26
shogun-buildbot_build #816 of deb2 - static_interfaces is complete: Success [build successful]  Build details are at
shogun-buildbot_build #359 of rpm1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Soeren Sonnenburg <>10:58
blackburnsonney2k: I finally managed to wake up11:22
blackburnsonney2k: so we can release?11:23
@sonney2kblackburn, I think so11:58
@sonney2kblackburn, and good morning russia :)11:58
blackburnsonney2k: heh I was having fun with cmake whole night so it is not morning already11:59
* sonney2k runs make release11:59
blackburnokay I've got to write a mail11:59
blackburnsonney2k: is it late to update news?12:00
@sonney2kmake data-release fails :/12:00
@sonney2kblackburn, ohh what do you want to do?12:00
blackburnsonney2k: I forgot to mention that last bug12:01
blackburnsonney2k: ah it is already here hah12:01
blackburnsonney2k: I tried to fix that, but failed to do that right12:01
@sonney2kblackburn, oh #*$#YH!12:01
blackburnsonney2k: what's up?12:01
@sonney2kwe forgot to increase data counter12:01
@sonney2kin NEWS12:01
@sonney2kblackburn, so add whatever you need to NEWS12:02
@sonney2kand change data to 0.512:02
blackburnsonney2k: what about parameter version?12:02
@sonney2kI would say doesn't matter12:03
@sonney2kwe screwed that anyways12:03
@sonney2kohh and please mention that bug I hopefully fixed12:03
@sonney2kabout sg('lclassify') returning -1/+1 labels12:03
blackburngot it12:03
shogun-buildbot_build #360 of rpm1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Soeren Sonnenburg <>12:04
blackburnsonney2k: done12:05
shogun-notifier-shogun: Sergey Lisitsyn :master * a0e0d95 / src/NEWS:
shogun-notifier-shogun: Updated NEWS12:05
@sonney2kblackburn, SGString is not based on SGReferencedData12:11
@sonney2knews is still telling this12:12
blackburnsonney2k: yes12:12
blackburnsonney2k: string list is though12:12
shogun-notifier-shogun: Sergey Lisitsyn :master * 89d073b / src/NEWS:
shogun-notifier-shogun: SGString is not based on SGReferencedData12:13
@sonney2kblackburn, ohh and the date is 03-17 :)12:13
@sonney2kblackburn, do you fix it?12:14
blackburnsonney2k: yes12:14
shogun-notifier-shogun: Sergey Lisitsyn :master * 9702994 / src/NEWS:
@sonney2kor shall I12:14
shogun-notifier-shogun: Fixed date of release12:14
* sonney2k runs make-prepare-release again12:15
@sonney2kblackburn, look at this...
@sonney2ksomething weird is going on12:15
blackburnsonney2k: I am quite sure it is a heisenbug12:16
blackburnsonney2k: all libshogun examples pass on my machine - so we play with probabilities haha12:16
@sonney2kblackburn, it is real on this machine - fails for some time now...12:21
@sonney2kno idea tough12:21
@sonney2kit works everywhere else12:21
@sonney2kblackburn, anyway prepare the announcement email in an hour so so we should be ready12:21
blackburnsonney2k: alright12:21
@sonney2kblackburn, argh _ don't really know how to properly udpate the website12:29
blackburnsonney2k: what do you mean?12:30
@sonney2kpreviously make release did fully take care of this12:30
@sonney2kmake update-website12:30
@sonney2kbut now NEWS is not automatically parsed :/12:30
-!- heiko [] has joined #shogun12:30
@sonney2kblackburn, I need n4nd0 I guess or bern4rd12:31
blackburnsonney2k: asked him to join #shogun12:31
@sonney2kheiko, soon soon12:32
heikosonney2k: release? :)12:32
@sonney2kheiko, yeah - we would have one if NEWS was upto-date12:32
-!- n4nd0 [] has joined #shogun12:32
heikosonney2k:  blackburn fixed another bug yesterday, that could go in there12:33
@sonney2kheiko, already in12:33
@sonney2kheiko, release date was wrong too :)12:33
@sonney2kn4nd0, I am lost again with the website12:34
heikosonney2k: haha :)12:34
@sonney2kn4nd0, we need to change it to 2.1 and NEWS too12:34
n4nd0sonney2k: what?12:34
n4nd0was does it imply to change it to 2.1 and what do you want to change of the NEWS?12:34
@sonney2kn4nd0, we have that new version available and 2.0 on the website12:39
@sonney2kI am trying to figure out how to cahgne it to 2.112:40
@sonney2kthe news file I figured out by now12:40
@sonney2kn4nd0, sigh it is an image!12:43
n4nd0sonney2k: like everything in the webpage :D12:44
@sonney2kn4nd0, except the text :D12:44
@sonney2kn4nd0, now I would wish I knew the font12:45
@sonney2kblackburn, let us introduce version subtiles12:47
blackburnsonney2k: subtiles?12:47
@sonney2klike 'mighty iceberg'12:47
@sonney2kany suggestions?12:48
blackburnsonney2k: borsch12:48
-!- heiko [] has quit [Quit: Leaving.]12:48
blackburnsonney2k: vodka12:48
@sonney2kblackburn, I don't know if "In vodka we trust" will gain us users :D12:49
@sonney2kwiking, n4nd0, what do you guys drink over there12:49
@sonney2kblackburn, "Powered by Vodka, Beer and Mate"12:50
blackburnsonney2k: girls and vodka12:50
blackburnsonney2k: if I didn't make any mistake there were ~8900 commits since last release12:53
blackburnI am afraid I did12:54
blackburnyeha 9705 totally12:55
blackburnI got wrong12:55
@sonney2kblackburn, not good12:57
-!- sidi_ [~sidi@] has quit [Read error: Connection reset by peer]12:57
@sonney2ktoo much between releases12:57
blackburnsonney2k: not good what?12:57
@sonney2klets release more often12:57
blackburnsonney2k: I did count that to impress people12:58
wikingsonney2k: palinka12:58
wikingsonney2k: mmm do we have tag?12:59
wikingblackburn: which one is the release commit?12:59
blackburnwiking: palyonka in russian is a slang name for pirate vodka made of some shit like acetone12:59
@sonney2kblackburn, wiking n4nd0 - current subtitle is (for real!) Powered by Vodka, Beer and Mate.12:59
@sonney2kI could still exchange beer (as I dont' drink any)13:00
blackburnwiking: I don't know, put any commit made on 1st of sept13:00
wikingand 2.1?13:00
blackburnwiking: I hope 2.1 will be tagged now13:00
wikinglets see13:01
wikingwhen is it coming? :)13:01
* sonney2k is uploading the release13:01
wikingok so then we have a commit version already13:01
@sonney2k(make update-webpage)13:01
wikingsonney2k: can u copy-paste here13:01
@sonney2k$ git branch13:01
@sonney2k  master13:01
@sonney2k* shogun_2.113:01
wikingis that pushed?13:02
wikingseems not yet13:02
@sonney2kgit tag | grep shogun_2.13:02
@sonney2kwiking, not yet13:02
wikingok so shall we create a dev branch?13:02
@sonney2kuploading first13:02
wikingcalled dev or development13:03
@sonney2kyes dev is long enough13:03
wikinglet's go with that then13:03
wikingso from now on all the stuff go into dev13:03
wikingand keep master as stable..?13:04
wikingor the other way round13:04
wikingkeep master as development branch13:04
wikingand create a stable?13:04
@sonney2kblackburn, do you recall that email from ?#?$# on the mailinglist about proper git development?13:04
wikinganyhow the default branch on github should be the stable branch13:04
blackburnsonney2k: yes13:04
@sonney2khe suggested a scheme that we figured woudl be very useful13:04
@sonney2kI cannot find it13:05
-!- sidi_ [~sidi@] has joined #shogun13:05
@sonney2k(as usual)13:05
blackburnsonney2k: he was mentioned in news I think13:05
blackburnsonney2k: evan shelhamer13:05
wikingit's with git flow13:06
wikingpretty handy tool13:06
wikingwe could start using that13:07
shogun-buildbot_build #361 of rpm1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Sergey Lisitsyn <>13:18
wikingok dev branch is there13:24
wikingi'll try start using git-flow13:25
blackburnsonney2k: writing mails is pain sometimes haha13:36
@sonney2kblackburn, releases etc should be uploaded13:37
@sonney2kdoc is still uploading13:37
blackburnsonney2k: mail is not yet ready anyway13:37
@sonney2kdoc is still uploading though13:42
wikingwith git flow13:45
wikingbut afaik in this case PR-ing our code is useless13:47
wikingi mean if we choose to go with git-flow13:48
wikingthen we should directly work on shogun.git13:48
wikingbecause in that case PR is an unecessary overhead13:48
wikingi mean this for the developers who has write access to the shogun.git13:48
wikingas fetching around from other developers' repos features13:49
wikingis really just a painful overhead13:49
blackburnwiking: I do believe PRs for core devs is overhead13:51
shogun-notifier-shogun: Soeren Sonnenburg :master * 6ce4431 / src/ChangeLog:
shogun-notifier-shogun: updated changelog13:51
shogun-notifier-shogun: Soeren Sonnenburg :master * 1b1130d / doc/Doxyfile:
shogun-notifier-shogun: Preparing for new Release shogun_2.1.013:51
shogun-notifier-shogun: Soeren Sonnenburg :master * e6138dd / src/ChangeLog:
shogun-notifier-shogun: Preparing for new Release shogun_2.1.013:51
shogun-notifier-shogun: Soeren Sonnenburg :shogun_2.1 * 96e8031 / src/shogun/lib/versionstring.h:
shogun-notifier-shogun: Tagging shogun_2.1.0 release13:51
shogun-notifier-shogun: Soeren Sonnenburg :shogun_2.1 * 463c42b / doc/ (5 files):
shogun-notifier-shogun: update example list13:51
@sonney2kblackburn, any update on the release email?13:54
blackburnsonney2k: just a few minutes13:54
blackburnsonney2k: want to read it before?13:57
@sonney2kblackburn, send it13:57
wikingemail sent13:58
wikingbtw: sorry for my shitty explanation, but i'm a bit hangover :P14:00
blackburnwiking: palinka?14:02
blackburnsonney2k: n4nd0: wiking: see the mail14:02
@sonney2kannouncement is on mloss.org14:03
@sonney2kn4nd0, please twitter about the release!14:03
blackburnsonney2k: I'll do14:03
wikingblackburn: yes :)14:03
wikingblackburn: where's the email14:04
blackburnwiking: sent minute ago14:04
wikingon the list?14:05
blackburnwiking: I am afraid it was send by russian post14:05
@sonney2kblackburn, didn't see it14:05
blackburnhaha alright lets wait14:05
@sonney2kblackburn, did you sent it out?14:06
blackburnsonney2k: yes14:06
@sonney2kblackburn, to which address?14:06
blackburnsonney2k: I just forwarded it to you14:06
blackburnsonney2k: did you receive?14:07
blackburnsonney2k: may be something is wrong with address14:08
@sonney2kblackburn, yes arrived14:08
@sonney2koscar winning email :D14:08
blackburnsonney2k: a glitch of mailing list?14:08
n4nd0sonney2k: will tweet and put something in g+14:09
n4nd0when it appears in the webpage14:09
-!- votjak [] has left #shogun []14:09
blackburnn4nd0: ok let you do it :)14:09
blackburnsonney2k: I am afraid - what should I do, re-send?14:12
@sonney2kblackburn, relax and wait?14:14
@sonney2kblackburn, I send some news blob to freshmeat too14:14
blackburnsonney2k: ok lets wait then14:17
@sonney2kposted to G+14:18
@sonney2kposted to blog and thus debian planet14:22
shogun-buildbot_build #362 of rpm1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Sergey Lisitsyn <>14:22
wikingsonney2k: if we start using dev branch, shouldn't we switch buildbots to use that?14:23
@sonney2kwiking, we need to discuss how we do things14:23
blackburnsonney2k: I could send it again may be14:23
@sonney2kmaybe this evening?14:23
wikingany idea when heiko will be around again?14:24
@sonney2ktoo bad he left14:24
@sonney2kI figured some issues in the relese Makefile14:26
@sonney2kwhere how do I fix this?14:26
@sonney2kwiking, btw is it properly tagged now?14:26
blackburnsonney2k: can you view that?14:27
wikingsonney2k: seems so14:29
@sonney2kblackburn, yes why?14:31
blackburnsonney2k: strange peak :D14:31
blackburnsonney2k: still don't understand what happened14:31
blackburnsonney2k: its tapkee downloads14:31
@sonney2kblackburn, is that tapkee?14:31
blackburnerr views14:31
@sonney2ksame date as ?14:32
blackburnsonney2k: near14:32
@sonney2kblackburn, wiking n4nd0 - so now we need to a) decide about future development process (maybe even with irc meetings from time to time!) b) workshop announcement / advertisement  c) gsoc planning14:34
blackburnsonney2k: next week is to *finish* proposal and ideas for me14:35
blackburnsonney2k: I'll let you all know how can you help14:35
blackburnwhere the heck the mail is14:41
shogun-buildbot_build #363 of rpm1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Soeren Sonnenburg <>15:29
-!- sidi_ [~sidi@] has quit [Read error: Connection reset by peer]16:07
-!- sidi_ [~sidi@] has joined #shogun16:13
-!- heiko [] has joined #shogun16:28
-!- n4nd0 [] has quit [Ping timeout: 264 seconds]16:33
heikosonney2k: blackburn, wiking, just saw the update! :)16:36
blackburnheiko: we are still waiting the mail to come16:36
heikoblackburn: who writes it? S?ren?16:36
blackburnheiko: I did and sent it..16:36
blackburnit arrived!16:37
heikoseeing it :)16:37
heikoblackburn: yeah! :) well done!16:39
-!- shogun-notifier- [] has quit [Quit: transmission timeout]16:51
wikingheiko: yo17:09
wikingheiko: check gitflow17:09
wikingheiko: and the mail i wrote about it17:09
-!- hoijui [] has quit [Ping timeout: 252 seconds]17:10
-!- n4nd0 [] has joined #shogun17:11
n4nd0blackburn: nice work with the mail!17:25
blackburnn4nd0: heh thanks17:26
n4nd0Shogun 2.1 powered by vodka, beer and mate haha17:26
blackburnn4nd0: it took a few hours for mail to arrive17:26
wikingheheeh yeah blackburn thnx for the mail! :)17:26
blackburnwiking: I am still failing to use clang analyzer haha17:29
-!- n4nd0 [] has quit [Quit: Reconnecting]17:29
-!- n4nd0 [] has joined #shogun17:30
blackburnwiking: that's what I always get:  scan-build: Removing directory '/tmp/scan-build-2013-03-17-1' because it contains no reports.17:30
-!- heiko [] has quit [Quit: Leaving.]18:58
-!- sidi [~sidi@] has joined #shogun19:14
-!- sidi_ [~sidi@] has quit [Read error: Connection reset by peer]19:14
-!- kishan [1883ef0b@gateway/web/freenode/ip.] has joined #shogun19:35
-!- heiko1 [] has joined #shogun19:44
-!- n4nd0 [] has quit [Ping timeout: 246 seconds]19:52
heiko1wiking, I read quickly and will do in more detail19:56
heiko1wiking, my main point is to have a sophisticated dev-process of some form (eg the one you proposed)19:57
heiko1and more importantly, hake tests madatory, and check whether tests pass before merging19:57
heiko1so that it is not possible to commit code which breaks tests/examples19:57
heiko1(at least on same basic level - not all configurations)19:57
heiko1and another thing I would like to have is auto-rejects19:58
blackburnheiko1: autorejects of?19:58
heiko1these should happen when tests fail /leak, warnings, missing documentation, incomplete examples, 0% test coverage19:58
heiko1blackburn:  of pull requests19:58
heiko1blackburn: so that we can only merge if things work19:58
blackburnheiko1: travis already does that19:59
heiko1there should be a hotfix branch to avoid this in emergencies, but in general it should not be possible to merge code which doesnt work19:59
heiko1blackburn: even the core-devs should not be allowed to merge19:59
heiko1blackburn:  currently, we only get the info, but still can press merge19:59
heiko1which is what we do19:59
heiko1should not be possible19:59
heiko1this way we force ourselves to do proper patches19:59
blackburnyou must have been in army :D20:00
heiko1blackburn: no, just talked to some friends who work in software engineering20:01
heiko1 :)20:01
heiko1if you merge a patch in a company and it doesnt work, you basically get killed :)20:01
blackburnheiko1: oh really? :D20:01
heiko1but also increases discipline (warnings/documentation) etc20:01
heiko1one can also do other cool things with such a process20:02
blackburnheiko1: I pulled in a crazy number of breakers during doing my job actually20:02
blackburnthat could be true for some companies but in general it is not such disciplined20:02
heiko1blackburn:  I know20:03
heiko1defining the ideal here20:03
heiko1blackburn:  one could to some other cool things with that20:03
blackburnheiko1: such discipline requirements should have a reason20:04
heiko1eg only accept examples if they come with a description and produce a screenshot20:04
heiko1blackburn: well I can argue20:04
heiko1so what happens before a release=?20:04
heiko1big bug hunt20:04
heiko1very annoying20:04
heiko1how are warnings fixed?20:04
heiko1by empty doxygen comments20:04
heiko1example descriptions? never done20:04
blackburnheiko1: that's true but there should be an explanation20:05
heiko1but we have to do these things *anyway*20:05
heiko1so why not immediately20:05
heiko1its actually less work if you do it when you write the code20:05
heiko1same with tests20:05
@sonney2kheiko1, blackburn, wiking - so whats going on?20:06
blackburnsonney2k: heiko1 wants to walk around with a rifle :D20:06
heiko1sonney2k:  we are raving about the release :)20:06
blackburnand shoot ya leg!20:06
heiko1blackburn: its not more work, we have to remove warnings anyway20:06
heiko1and its so annoying to do this before release20:07
blackburnheiko1: it is a question of motivation (for me too)20:07
blackburnheiko1: that's a great motivation that is needed to have such restricted thing20:07
heiko1its just a matter of getting used to it20:07
heiko1for me, it costs way more motivation to do all these pain in the a** things20:07
heiko1I just want to minimise the time spent on stupid things20:08
blackburnheiko1: I am still lost with what do we do20:08
@sonney2kheiko1, blackburn alright lets discuss this20:08
blackburnsonney2k: discussion time?20:08
@sonney2kheiko1, the biggest issue I see is that people don'tlike to write doxygen comments20:08
heiko1sonney2k: true20:09
@sonney2klook at wiking - his CCSOSVM is still not well documented20:09
@sonney2kand we are admittedly no better20:09
blackburnlook at me, my code is not documented at all :D20:09
heiko1sonney2k: what is the point of writing code if you dont tell people how to use it?20:09
@sonney2kblackburn, documenting code is not what I meant20:09
heiko1sonney2k: agreed20:09
@sonney2kI meant API doc20:09
heiko1writing documentation is annoying20:09
heiko1thats why we should have a process that ensures it is done anyway20:09
heiko1sonney2k: but documentation is not my main concern in fact20:10
@sonney2kyes true but heiko1 we cannot do it with a black / white approach20:10
blackburnheiko1: the only way i see is metrics computation20:10
@sonney2kwe have to find something we can all somehow live with20:10
heiko1sonney2k: agreed, but I totally dont like this postponing forever20:10
blackburnheiko1: we should know weak points20:10
heiko1sonney2k: my biggest point actually is the fact that we loose code20:10
blackburnwe don't know them now20:10
@sonney2kheiko1, well one way would be to turn on warning -> error20:10
heiko1streaming for example20:10
heiko1json another one20:10
blackburnsonney2k: warnings are rather useless20:11
@sonney2kwe all the time fix warnings before a release be it doxygen or in code20:11
@sonney2kblackburn, they helped me a lot20:11
heiko1sonney2k, blackburn thing is we have to fix this stuff anyway20:11
@sonney2kheiko1, yes sure I agree ... but if we had integration tests turned on from 1.1 on we would have catched this20:11
heiko1sonney2k:  yes20:11
heiko1so make it madatory20:12
@sonney2kso my take on this is to *gradually* improve the situation20:12
heiko1then these things dont happen20:12
@sonney2ki.e. step 1) make integration test pass mandatory20:12
heiko1sonney2k: yeah of course, as said, I describe the "ideal situation"20:12
heiko1sonney2k: good idea.20:12
blackburnsonney2k: heiko1: I still want to define what do we develop20:12
@sonney2k2) force people to *repair* the tests20:12
@sonney2kheiko1, ^2 is what we also sometimes suck with20:12
@sonney2kwe had tests failing for 1 week or so20:13
heiko1sonney2k: this is another things which *has* to be done, so why not forbid to merge without passing tests?20:13
@sonney2kand heiko1 there is no way to force people to do it before submitting - otherwise you would have to compile *all* interfaces20:13
@sonney2kand need all shogun deps & compile time is like 1hr20:13
heiko1sonney2k: good point, so lets just do it for libshogun20:13
heiko1no interfaces20:13
heiko1main thing is the unit-tests should pass20:13
@sonney2kheiko1, libshogun is already taken care of by travis20:14
heiko1sonney2k: yeah, but Id like to change the ability to merge20:14
@sonney2kheiko1, but we usually don't struggle with libshogun stuff but the more involved code we have in all the *modular interfaces20:14
@sonney2k examples20:14
heiko1sonney2k: this will change20:14
heiko1unit tests are way more important for us20:14
heiko1examples catch errors on the very last bit of the code20:15
@sonney2kheiko1, ok I don't mind having libshogun unit tests to pass as a precondition to git commit20:15
heiko1this is not sufficient for fixing things20:15
heiko1unit tests (which should be madatory for new things) will  be able to tell you exactly what fails20:15
heiko1and I agree, building everything takes too long, so yeah this sounds good20:15
@sonney2kheiko1, so we could add a git pre-commit hook like here
heiko1let me have a look20:16
@sonney2kheiko1, so lets implement these 2 things for now - that should already help *a lot*20:16
heiko1sonney2k:  yes20:16
heiko1sonney2k:  I would also like to have a rejection if one has 0% unit-test coverage on code one commits20:17
heiko1this forces at least to think about one test20:17
heiko1we dont need a vast coverage20:17
@sonney2kheiko1, no idea how to do this20:17
heiko1sonney2k: me neither, more talking about the idea20:17
@sonney2kcomputing code coverage I mean20:17
heiko1sonney2k:  its probably possible to compute this, I have seen it in the gnunet project20:17
heiko1they use buildbot too and have this, I could ask them20:18
heiko1sonney2k:  what do you think of the general idea?20:18
blackburnheiko1: I am not sure that would work for us20:18
heiko1blackburn:  what exactly?20:19
blackburnheiko1: computing test coverage on commit..20:19
heiko1blackburn: you mean techinally?20:19
blackburnheiko1: yeah20:20
@sonney2kheiko1, yeah why not ask them20:20
@sonney2kanyway I think this is good enough for the start: the other big change would be the way we develop20:20
heiko1sonney2k: yeah gradually change things20:21
@sonney2kand I don't really know how we can do this20:21
@sonney2kif we introduce a dev and feature branches etc20:21
@sonney2kwhat does the buildbot build20:21
-!- kishan [1883ef0b@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]20:22
heiko1sonney2k, do we really need the feature branches?20:23
heiko1since small number of developers20:23
@sonney2kheiko1, sometimes I would think so - e.g. some big underlying change like we had with SGVector20:24
blackburnheiko1: would be better if you develop not through PRs but your own branch20:24
blackburn(I think so)20:24
heiko1ok agreed20:24
@sonney2kheiko1, blackburn - what is the minimal thing we want?20:25
@sonney2ka dev branch and a stable master?20:25
heiko1sonney2k:  yes20:25
heiko1they should have slightly different tests I guess20:26
heiko1stable should be compiled with warnings as errors for example20:26
blackburnsonney2k: master is stable, and feature branches20:26
heiko1blackburn:  general dev branch would be good20:26
heiko1master should be "production ready"20:26
blackburnmaster (stable) development and feature branches20:26
heiko1so our current release is master20:26
blackburnI wish we could allow commits to specific branches20:27
blackburnlike to feature branch but not master20:27
@sonney2kjust FYI - if we create a release we do create a release branch20:27
@sonney2kand can put in bugfixes in there20:27
heiko1sonney2k:  good idea20:27
@sonney2kthere is also one particular issue20:28
@sonney2kheiko1, that already is like this!20:28
@sonney2kwe autogenerate the lib/versionstring.h file20:28
heiko1sonney2k:  I see20:28
@sonney2kthat file is generated on configure/make20:28
@sonney2kand contains the current git version20:28
heiko1sonney2k: ok understoof20:28
@sonney2kwe use this git version when people send us bug reports20:28
heiko1what would be really good if we had a stable branch that is always in the condition that we could release20:29
@sonney2kwe cannot generate this file in a release shogun version20:29
heiko1with just one click20:29
@sonney2kheiko1, it cannot be 1 click20:29
@sonney2kheiko1, the release script is already very simple20:29
heiko1sonney2k: not to be taken literal20:29
heiko1sonney2k: I mean more the code is in the condition to release20:29
@sonney2kheiko1, look at the toplevel makefile20:30
heiko1we dont have to spend days in fixing things like removing warnings20:30
@sonney2k# * To make a release (and tag it) run20:30
@sonney2k#       make prepare-release20:30
@sonney2k#       make git-tag-release20:30
@sonney2k#       make release20:30
@sonney2k#       make data-release20:30
@sonney2k#       make update-webpage20:30
heiko1sonney2k: ok cool didnt know that20:30
blackburnha, mikio twitted about shogun20:30
@sonney2kbut then you will need to a) send out announcement to mailinglist, g+, twitter,, freshmeat, my blog20:30
@sonney2kblackburn, link?20:31
blackburn sonney2k20:31
heiko1sonney2k: will write a tweet later today20:31
heiko1sonney2k: ok these are non-automatic things20:31
blackburnheiko1: twitt is already here if you mean @ShogunToolbox20:31
@sonney2kheiko1, and the biggest concern is make prepare-release to actually go through20:31
@sonney2kheiko1, yeah n4nd0 did it :)20:32
@sonney2kheiko1, make prepare-release can take ~1 hour20:32
@sonney2kit compiles all interfaces / updates doc / runs all examples & tests20:33
@sonney2kthe rest is fast but can take 30 mins too just to upload data etc20:33
heiko1ok, so this is already there then20:33
heiko1question, how to make sure that we dont have this huge pile of work before release?20:34
@sonney2kheiko1, more or less ... there were some bugs in the scripts I would like to fix20:34
@sonney2kheiko1, release more often!20:34
@sonney2kheiko1, we have 6-12 months between releases20:34
@sonney2kno wonder that lots of stuff breaks20:34
heiko1how do releases prevent things from breaking?20:34
@sonney2kif we release every 1-2 months not so many features enter20:34
@sonney2k-> less testing needed20:34
@sonney2k-> faster release20:35
heiko1sonney2k:  we dont do testing currently20:35
@sonney2kheiko1, impossible even integration / unit tests won't help20:35
@sonney2kheiko1, I do20:35
heiko1sonney2k: I mean we have no systematic way of doing it20:35
@sonney2kheiko1, I run some example code and check if things are still ok20:35
@sonney2kheiko1, is there any?20:35
blackburnheiko1: you are a pure german ;)20:36
heiko1in my eyes this problem can only be solved if writing unit/integration is part of the development process20:36
heiko1we currently have no mechanism to ensure that things work20:36
heiko1we only detect if things crash20:36
heiko1so catching problems earlier I think would help a lot20:37
heiko1if one realised that an example crashes its usually too late20:37
heiko1take the GP stuff20:37
heiko1was implemented - worked fine20:37
@sonney2kheiko1, let me tell you that 100% unit / integration test coverage won't help you20:37
@sonney2k(from experience)20:38
heiko1then was extended  - which broke the first working part20:38
heiko1but no detection20:38
heiko1later on detected in example by pure chance20:38
heiko1sonney2k:  I know20:38
@sonney2kyou change sth and voila your cool unit/integration test exactly did not test for that borderline thing20:38
heiko1sonney2k: but *having* tests helps20:38
@sonney2khappens to me all the time20:38
@sonney2kheiko1, yes sure20:38
heiko1sonney2k: you are right, I am not saying this solves everything20:38
heiko1but my experience is that almost all bugs I fix can be catched by unit testing20:39
heiko1the multiclass SVm bug yesterday another example20:39
heiko1if there was a test that actually *ran* the code once, we would not have had this bug20:39
blackburnI tell that each minute but - do our users care so much about bugs? that's the question20:39
blackburnif we are striving for quality - what for?20:39
heiko1blackburn:  if you install shogun, and try an example, and all you get is sefault, you instantly delete20:40
@sonney2kheiko1, but we have to face reality too - if we don't want any future contributions we require 100% unit tests & 100% dyxgen api doc20:40
@sonney2kwe need some tradeoff20:40
heiko1not 100%20:40
@sonney2kfun <-> stability20:40
heiko1but not zero %20:40
heiko1thats my point20:40
@sonney2kheiko1, yes sure!20:40
heiko1at least think of *some* things to test20:40
heiko1that was my point, forbid 0% coverage20:40
-!- sidi [~sidi@] has quit [Read error: Connection reset by peer]20:40
heiko1not aim for 10020:40
heiko1we dont need that20:40
heiko1but at least *think* of it20:40
heiko1which we currently dont do20:40
heiko1especially important for the gsoc project to be suistainable20:41
blackburnheiko1: then some verbal pushes is what we can do20:41
@sonney2kblackburn, heiko1 - the most important thing is to nicely respond to bug reports on the mailing list - and I mean not like - ahh don't use that interface or I look at this when I find some time in a few weeks20:41
@sonney2kjust fix the bug and notify the user that it is fixed20:41
@sonney2kheiko1, yeah but it was our fault not requiring unit tests back then20:42
@sonney2kheiko1, but we were in no shape to have that a few years back20:42
@sonney2kheiko1, situation of shogun gradually improved over years20:42
blackburnsonney2k: yeah we weren't that wise we are now20:42
blackburnI feel ok about that20:42
heiko1my gf just started a fire in the kitchen20:42
@sonney2kheiko1, look - we had no unit tests not build bots no integration test no doc no nothing back then20:42
heiko1sonney2k: yes, we are on a good way20:43
blackburnheiko1: yikes20:43
@sonney2kheiko1, hand her some fire extinguisher20:43
@sonney2kthen continue chatting20:43
heiko1actually the l?schdecke20:43
heiko1haha :)20:43
blackburnoh these german words make me crazy20:43
@sonney2kwe just have pretty well buildbot infrastructure since last gsoc20:44
@sonney2kthis gsoc we will have integration / unit tests20:44
heiko1sonney2k: just want to push this test thing, I think we have to make ourselves change our habits20:44
@sonney2kand we will require students to write both or fail them20:44
heiko1sonney2k: yes thats really good20:44
heiko1exactly thats my goal :)20:44
@sonney2kbut we need to tell them upfront!20:44
heiko1sonney2k: yes, that was the point of the guidelines20:45
@sonney2kheiko1, we have README.developer in src20:45
@sonney2klets add the part about unit/integration tests20:45
heiko1good, lets merge it20:45
@sonney2knot all of it20:45
heiko1maybe after some reviewing20:45
heiko1sonney2k: what do you think of making examples fail if they do not return things or do not create a screenshot?20:46
blackburnsonney2k: heiko1: a good thing would be - all PRs should be 'ok' from all committers20:46
@sonney2kwe are not there yet20:46
heiko1blackburn: that would be good, but too much work20:46
blackburnheiko1: why?20:46
@sonney2kblackburn, everyone wouldhave to do code reviews -> impossible20:47
heiko1rahter make the system do as much as possible20:47
blackburnsonney2k: no I mean we three20:47
@sonney2kthat doesn't scale20:47
@sonney2kone of us20:47
blackburnI don't like that - there should be a few reviewers20:47
@sonney2kand if sth breaks we punish the merging guy :D20:47
heiko1student patches should not be merged if anything is missing :)20:48
heiko1guys, as you probably already thought, diner is ready soon here,20:48
heiko1its roast :)20:48
heiko1I gotta buy some things and then eat.20:48
heiko1so sorry, gotta go20:48
blackburnheh alright20:48
@sonney2kheiko1, blackburn btw are you ok with the slogan on the website?20:48
blackburnsonney2k: hah sure20:48
heiko1sonney2k: sure :)20:49
blackburnsonney2k: I like crazy things - being serious is a fault20:49
heiko1sonney2k: you should add coffee and almonds for my part :)20:49
-!- heiko1 [] has quit [Quit: Leaving.]20:49
@sonney2kheiko1, chocolate here20:49
@sonney2kblackburn, so I guess we agree on the unit test stuff but how do we do it with the git branches etc?20:50
blackburnsonney2k: good q20:50
@sonney2kblackburn, it is really pressing for me to fix some erros in the release scripts while I still remember what they were20:50
@sonney2kI am totally clueless what the buildbot should then check20:51
blackburnsonney2k: both..20:51
blackburndevelopment and stable20:51
blackburnsonney2k: need advice20:53
@sonney2kblackburn, and feature branches?20:54
blackburn sonney2k plots, page 320:54
blackburnsonney2k: no, feature branches should not be builded I think20:54
-!- sumit [73f91219@gateway/web/freenode/ip.] has joined #shogun20:55
@sonney2kblackburn, IIRC there is subfix or sth where you can have >1 fig20:55
blackburnsonney2k: I don't get it - what is it for?20:56
@sonney2kto have the figures name a) b) c)20:56
blackburnsonney2k: ahh20:57
blackburnsonney2k: but figures? are they ok?20:57
@sonney2kblackburn, btw I like evan shelhamers suggestion20:57
@sonney2kblackburn, yes20:57
@sonney2kto create a release branch to stabilize things20:57
blackburnsonney2k: I don't know what to do with faces though - they are not visible, too small..20:58
@sonney2ka long time before we release20:58
blackburnsonney2k: yes that's a good way20:58
@sonney2kso we could even do sth like this:20:58
blackburnsonney2k: kernel do the same?20:58
@sonney2krelease every month20:58
@sonney2kwith a delay of one month20:58
@sonney2kwe just create a shogun_2.2 branch in one month20:58
@sonney2kthen fix things for 1 month and release20:59
@sonney2kand at the same time develop in master20:59
@sonney2kdev branch20:59
blackburnsonney2k: looks good but don't know how well in practice20:59
@sonney2kblackburn, what I dont' understand though is how we do it all with github. I mean w/ github we have these pull requests21:00
@sonney2kwhich are branches itself right?21:00
@sonney2kso these are feature branches?21:00
@sonney2kwe intend to merge to the dev branch - correct?21:01
blackburnsonney2k: it is ok to PR to some branch21:01
blackburnsonney2k: why not to feature branch?21:01
@sonney2kblackburn, don't you have to create a branch to even do a PR?21:02
@sonney2kso isn't that then the same like a feature branch?21:02
blackburnsonney2k: well you can do a PR from a master21:02
blackburnsonney2k: selective rights would be awesome21:03
@sonney2kblackburn, you mean from your fork's master right?21:03
blackburnsonney2k: yes21:03
@sonney2kblackburn, yeah well ok but that kind of is a feature branch ...21:04
@sonney2kbut only a local one21:04
blackburnsonney2k: yes21:04
@sonney2kwe probably don't need shared feature branches (at least not often)21:04
blackburnsonney2k: it is not really relevant for us actually21:04
@sonney2kso all that we want to change is have a dev branch & a master branch21:05
blackburnsonney2k: yeah I think dev+master would be enough for now21:05
@sonney2kthe only problem I have is that we cannot git tag the released version in the master branch21:05
@sonney2kbecause we include lib/versionstring.h in the release21:05
@sonney2kwhich we *never* put in the git repo in master/dev21:06
@sonney2kblackburn, btw our shogun2.1 announcement has 14 +1's on g+ & 3 reshares21:07
blackburnsonney2k: quite cool21:07
@sonney2kblackburn, wiking so what do we do? we cannot git tag in dev/master but have to do it in a release branch21:08
@sonney2kcan we live with that?21:08
blackburnsonney2k: I don't know21:09
@sonney2kblackburn, anyway no matter what - let me create a dev branch and then I can commit my fixes there21:13
blackburnsonney2k: sure21:14
@sonney2kblackburn, and you do the gsoc stuff otherwise we won't make it21:14
blackburnsonney2k: yeah I'll spend some time tonight on that21:14
blackburnsonney2k: so I should not expect real help from mentors, right?21:17
blackburnsonney2k: I asked cheng to write some descriptions but no answer at all :D21:18
@sonney2kblackburn, no you should get some push them as much as you can but prepare at least 1 example to give them something they can base their stuff on!21:18
blackburnsonney2k: I gave them 2012 ideas21:18
@sonney2kblackburn, I see you created a dev branch on github21:18
blackburnsonney2k: I?21:18
@sonney2kblackburn, how do I get that pulled?21:18
@sonney2kI don't see it locally not with git branch -a (as remote branch)21:19
blackburnsonney2k: I don't get the question21:19
blackburnsonney2k: did you create your branch?21:19
@sonney2kblackburn, no21:19
blackburndo that21:19
@sonney2kat github there is a dev branch21:20
blackburnsonney2k: you wanted a branch?21:20
@sonney2kfor shogun21:20
blackburnsonney2k: where? :D21:20
@sonney2kLast updated 9 hours ago by lisitsyn.21:20
blackburnsonney2k: ehmmm - where do you see it?21:21
@sonney2kblackburn, I just type dev into branches21:22
blackburnsonney2k: the thing is that I didn't create any branch21:22
blackburnI just pulled and what I saw is21:22
blackburn   9702994..e6138dd  master     -> origin/master21:22
blackburn * [new branch]      dev        -> origin/dev21:22
@sonney2kit is there now!?21:23
blackburnsonney2k: wiking may be?21:24
wikingjust a sec21:24
wikingi just came back21:24
@sonney2kso how do I work on this?21:24
blackburnwiking: is that you who created 'dev'?21:24
@sonney2kit is a remote branch so I can only push to this right?21:25
wikingi even told here21:25
blackburnwiking: yeah21:25
wikingcheck logs21:25
wikingabout 5 hours ago i told21:25
blackburnwiking: I remember something :)21:25
wikingthat there's dev branch now in git21:25
@sonney2kwiking, you did this before we released so we need to rebase it to master21:25
@sonney2kso I create a dev branch locally based on that?21:25
wikingwe need a simple repbase21:25
@sonney2kand then git rebase21:25
@sonney2kand push to the dev branch?21:25
wikingnaaah just a sec21:25
wikingi'll fish it21:25
wikingfix it21:26
@sonney2kno no21:26
@sonney2kI want to do it21:26
@sonney2kand learn how to :)21:26
wikingwith git flow?21:26
wikingbtw do we agree to use that?21:26
-!- sumit [73f91219@gateway/web/freenode/ip.] has quit [Quit: Page closed]21:26
@sonney2kwiking, no21:26
wikingso no gitflow?21:26
wikingno direct writing to git shogun repo?21:26
wikingand PRing every stuff?21:27
@sonney2kwiking, no idea but likely not git flow21:28
wikingi'm totally 4 gitflow :)))21:30
wikinghelps a lot in branch maintaining21:30
wikingespecially the whole release - dev branch maintaining21:31
@sonney2kwiking, we have the issue with our lib/versionstring.h not being part of git master/dev21:33
wikingyep i get it21:34
@sonney2kwiking, no idea what to do about it21:35
wikingjust a sec21:37
@sonney2kwe have to think about it a bit more21:38
wikingsonney2k: what's against having lib/versionstring.h in the repo?21:49
wikingi mean we can even have like a template version of that in the repo, and have a makefile script that automagically updates it? :P21:50
wikingand that's just part of the actual release....?21:50
wikinghere's an example:
-!- hoijui [] has joined #shogun22:00
-!- n4nd0 [] has joined #shogun22:24
-!- zxtx [] has quit [Ping timeout: 246 seconds]22:48
-!- n4nd0 [] has quit [Quit: leaving]22:51
wikingsonney2k: so actually regardless of gitflow23:04
wikingsonney2k: can we agree that if one has a write access on shogun-toolbox/shogun.git repot then he can push commits directly into dev branch w/o doint the PR->merge cycle?23:05
-!- hoijui [] has quit [Ping timeout: 252 seconds]23:43
--- Log closed Mon Mar 18 00:00:14 2013