Open in new window / Try shogun cloud
--- Log opened Fri Jan 10 00:00:45 2014
-!- besser82 [quassel@fedora/besser82] has quit [Remote host closed the connection]01:56
-!- besser82 [quassel@fedora/besser82] has joined #shogun02:20
shogun-buildbotbuild #676 of nightly_default is complete: Failure [failed test]  Build details are at
-!- new_lido [~walid@] has joined #shogun05:47
-!- new_lido [~walid@] has quit [Quit: Leaving]05:57
-!- sonne|osx [~sonne@] has joined #shogun08:09
-!- sonne|osx [~sonne@] has quit [Quit: sonne|osx]08:27
besser82sonne|work:  the problem isn't the pickle-protokoll09:56
besser82sonne|work:  it is the incompat of binary-structures09:56
besser82sonne|work:  like one can load pickles from Py2.X in Py309:56
besser82sonne|work:  an restore their contents...09:57
besser82sonne|work:  but not that is not guranteed for c-compiled extension09:57
besser82sonne|work:  basic types aren't the prob (str, int, float, list,...)09:58
besser82sonne|work:  they will surely work09:58
besser82sonne|work:  but stuff like numpy-array isn't compat09:58
besser82sonne|work:  it can be restored with a trick, but the contents is garbled09:59
besser82sonne|work:  and instances of classes from a c-compiled ext. are likely to fail / segfault09:59
besser82sonney2k:  ^^^09:59
besser82sonne|work, sonney2k:  but that actually is in nature of those pickles10:03
besser82sonne|work, sonney2k:  but I just found a "nasty" hack that may or may not help; need to try that10:04
besser82sonne|work, sonney2k:  but I just found a "nasty" hack that may or may not help; need to try that10:04
besser82sonne|work, sonney2k:  we possibly might be able to restore the pickles by aletring them "manually" to reflect pickled str as bytes;  that might do the trick...10:05
besser82sonne|work:  I assume the problem of that segfault is caused by sth. like the incompat of Py3::str and std::string / cstr10:20
besser82sonne|work:  cstr == char*10:20
besser82sonne|work:  a Py3::str may have a memory footprint of mixed width dependending on the type of char in it10:21
besser82sonne|work:  so some char may have a width of 2 or 4 bytes, others just 1 byte10:22
besser82sonne|work:  when std::string / char* is tied to 1 byte a char10:22
besser82sonne|work:  so passing a Py3::str-type (created from Py2::Pickle) into c / c++ as is may cause that segfault10:23
besser82sonne|work:  so I need to check whether this happenes when passing Py3::bytes to std::str / char*10:24
besser82sonne|work:  Py3::bytes should be perfectly char*10:25
besser82sonne|work:  which may / should10:25
besser82sonne|work:  do the trick...10:25
besser82sonne|work: same assumption for numpy-array  ;)10:26
sonne|workbesser82: yeah scary hack.10:26
sonne|workbesser82: we need some other way in the future10:27
besser82sonne|work:  other way for what? serialization of shogun::object?10:27
besser82sonne|work:  pickling / data exports10:27
-!- Saurabh7 [~Saurabh7@] has joined #shogun10:28
* besser82 throws an instace of PuzzledException: What() Which other way for what?10:29
-!- sonne|osx [] has joined #shogun10:34
-!- sonne|osx [] has quit [Ping timeout: 272 seconds]10:39
sonne|workbesser82: I think currently we compare objects with some string comparison or so - maybe we can just load them and have for different object types different comparators10:55
-!- thoralf [] has joined #shogun11:36
besser82sonne|work:  the assertions aren't the problem  ;)11:40
besser82sonne|work:  the problem is restoring the *pristine* contents of the pickles11:41
besser82thoralf:  Hey!11:41
thoralfHey Bjesser11:41
besser82sonne|work:  since the _Unpickler passes incompat binary-data into the modshogun / numpy classes  :-P11:42
besser82thoralf:  :D11:42
besser82sonne|work:  so we should change the way how objects are serialized for storage  ;)11:44
besser82sonne|work:  like using json instead of pickles  ;)11:45
besser82sonne|work:  which is pretty indep from lang / os / arch11:45
besser82sonne|work:  only carries nice ascii which then gets eval() and pushed into the data-structs11:46
-!- thoralf [] has quit [Ping timeout: 264 seconds]12:19
-!- thoralf [] has joined #shogun12:29
-!- Saurabh7 [~Saurabh7@] has quit [Quit: Leaving]12:48
sonne|workbesser82: ahh because it is no longer a ascii but whatever string right?12:56
-!- thoralf [] has quit [Ping timeout: 245 seconds]13:04
@wikingbesser82: there are miner issues with json...13:25
@wikingbesser82: see the serializationjson unit test13:25
-!- lisitsyn [] has joined #shogun13:36
besser82sonne|work:  nope, not that encoding issue...  It's da Py3-internal string-footprint...14:12
besser82wiking:  thx 4 info14:13
besser82sonne|work:  as I've written abouve14:13
-!- thoralf [] has joined #shogun14:22
-!- zxtx_ [] has quit [Ping timeout: 252 seconds]14:47
sonne|workbesser82: could you give me a minimal pointer how to do a spec file?15:17
sonne|workbesser82: I need to quickly create an rpm...15:17
-!- thoralf [] has quit [Ping timeout: 246 seconds]15:21
-!- zxtx_ [] has joined #shogun15:42
-!- bb_ [458fc8de@gateway/web/freenode/ip.] has joined #shogun15:52
-!- ChenZu [3ad5332a@gateway/web/freenode/ip.] has joined #shogun16:03
ChenZuAnybody can help me ? I'am crazy with shogun installation with matlab16:04
-!- lisitsyn [] has quit [Quit: Leaving.]16:04
ChenZuI execute the command 'sudo make install' and there no sg.mex64 anything in the matlab directory16:05
-!- ChenZu [3ad5332a@gateway/web/freenode/ip.] has quit [Quit: Page closed]16:21
besser82sonne|work:  still in need for rpm?17:00
besser82sonne|work:  rpmdev-newspec is creating some skel17:00
sonne|workbesser82: yeah still in need17:00
besser82sonne|work:  %prep %build %install %check %clean and such are like bash-script17:00
besser82sonne|work:  need some template?17:01
besser82sonne|work:  what kind of sw?17:01
sonne|workon monday17:01
besser82sonne|work:  kk17:01
sonne|workwork stuff17:01
sonne|workbut yeah cool...17:01
sonne|workbesser82: I basically have some directory17:01
sonne|workthat I just want to copy somewhere under /home/user17:02
sonne|workand ship this as rpm17:02
sonne|workthat is all17:02
besser82sonne|work:  then tar.gz that dir, list tarball as Source017:02
sonne|workbesser82: probably ultra easy for you17:02
besser82sonne|work:  yes, just a few seconds of work17:02
besser82sonne|work:  shall i giv e you template?17:03
besser82sonne|work:  like having some dummy tar and performing needed action of that17:03
sonne|workjust this tar line17:04
besser82first create tarball of the dir you want to ship: tar cf $my.tar $src17:05
besser82sonne|work:  ^^^  then compress17:05
besser82sonne|work:  like gz -9 $my.tar17:05
besser82sonne|work:  then create some spec17:05
besser82sonne|work:  %prep17:05
besser82sonne|work:  %setup -qn $dir_in_tar17:05
besser82sonne|work:  %build17:06
besser82sonne|work:  # noop17:06
besser82sonne|work:  %install17:06
besser82sonne|work:  mkdir -p %yourtarget && cp -a . $your_target17:06
besser82sonne|work:  %changelog17:06
besser82sonne|work: done17:06
sonne|workbesser82: sounds good17:08
sonne|workoki more monday17:10
* sonne|work off home17:10
besser82sonney2k:  I just prepared a rpm.spec example, doing what you said it should...17:44
besser82sonney2k:  shall I mail it to you?17:44
-!- iglesiasg [] has joined #shogun18:43
-!- mode/#shogun [+o iglesiasg] by ChanServ18:43
-!- FSCV [~FSCV@] has joined #shogun19:02
-!- iglesiasg [] has quit [Quit: Leaving]19:04
@sonney2kbesser82, yes please thanks!19:06
@sonney2kbesser82, I am already happy though19:06
-!- lisitsyn [~lisitsyn@] has joined #shogun19:39
@sonney2klisitsyn, how is your holidays?19:39
lisitsynsonney2k: had exam today :D19:40
lisitsynyes I can notebook a little19:40
@sonney2klisitsyn, I thought you had vacations?19:42
lisitsynsonney2k: they are over and exams are still here :(19:43
lisitsyngood thing about it is that is the last exams I have to take any longer19:43
lisitsynwow some major changes here19:45
@sonney2klisitsyn, parse error... that was your last exam?19:46
@sonney2klisitsyn, major changes?19:46
lisitsynsonney2k: haha okay lets do it again19:47
lisitsynsonney2k: I have two more exams but these are last as I'll just have to write up some thesis19:47
lisitsynsonney2k: major changes is what I see in git pull ;)19:47
lisitsynis that more parseable? ;)19:48
@sonney2klisitsyn, yes perfectly parseable :D19:48
@sonney2klisitsyn, not so big changes - only in the build system19:48
@sonney2klisitsyn, besser82 did a lot of python3 fixes19:49
@sonney2klisitsyn, and we now have a new buildbot that checks for python319:49
@sonney2klisitsyn, quite a lot already works19:49
-!- zxtx_ [] has quit [Ping timeout: 246 seconds]19:51
lisitsynsonney2k: cool19:53
lisitsynsonney2k: I am curious whether some quite major -> py3 conversion happens any day19:54
@sonney2klisitsyn, where?19:55
lisitsynsonney2k: I just mean it is like between 2.7 and 3.x still19:56
@sonney2klisitsyn, I would say all major extensions are ported by now19:56
@sonney2klisitsyn, doesn't mean it is used though - but I would guess that soon linux distributions are all switching over to python3.3 as default19:58
@sonney2ksoon as in their next stable release19:59
lisitsynsonney2k: makes sense20:01
* sonney2k is wrestling with the debian packages again...20:10
@sonney2klisitsyn, btw did you know we have 4 weeks until gsoc deadline20:11
besser82sonney2k, lisitsyn: for fedora transition to py3 as default is scheduled for f22 ~ Dec. 2014 ;)20:17
-!- bb_ [458fc8de@gateway/web/freenode/ip.] has quit [Ping timeout: 272 seconds]20:18
besser82sonney2k, lisitsyn: python2 basically is dead, by means of interp-development20:19
@sonney2kbesser82, yeah it just plain works(tm)20:19
besser82sonney2k, lisitsyn: 2.7.X is last release of py2-series; only bugfixes / security maintainance  ;)20:20
besser82sonney2k:  Works? what?  Py3?20:20
@sonney2kboth :)20:20
@sonney2kbut 2.7 is rock stable20:20
besser82sonney2k:  yes it is prooven stuff for sure; RHEL7 comes with it, but no Py3  :(20:21
adrinbesser82, don't worry about fedora deadlines, I don't remember the last time they were actually on time :D20:34
@sonney2kadrin, sounds like debian deadlines ;)20:35
besser82adrin:  I know  ;)  I'm one of those Fedora-guys  ;)20:36
adrinbesser82: hahahahaha20:37
besser82adrin:  :-P20:37
adrinbesser82: oh cool, I see :D pitty that I immigrated a while ago, unless I could blame you all the time :P20:39
besser82adrin:  :-P  Then just come back and checkout all brand-new bugz in new F2020:40
lisitsynsonney2k: 4 weeks?20:41
adrinbesser82: lol, I just found rolling distros more comfortable at some point, but thanks for the offer ;)20:42
besser82adrin:  fedora offer rolling, too  ;)  Checkout rawhide20:43
adrinbesser82: lol, yeah I remember it, I used to switch to rawhide all the time no later than a beta release. But you have to admit alphas and pre-alphas are really really unstable.20:44
besser82adrin:  not what I saw from the last 3 pre-alphas20:45
besser82adrin:  there had been some quirks for sure, but they were pretty good for workstation use20:45
adrinbesser82: I see, the last time I had a fedore was probably three releases ago, I'll have a look again :-)20:46
besser82adrin:  I had been shifting to F18 / F19 / F20 pre-alpha immediatly after branch-off and basically all stuff was usable, but some minor gliches  :D20:46
besser82adrin:  nice to hear!20:46
adrinbesser82: after all, I'm way more comfortable around rpm baseds.20:47
besser82adrin:  btw. there are nice shogun-rpms avail in official fedora-repo  ;)20:47
adrinbesser82: I'm sure glinches are not worst than gentoo ones ;)20:47
besser82adrin:  not even comparable  ;)20:47
@sonney2klisitsyn, I told you we are late this time with me not pushing you guys :D20:48
besser82adrin:  the worst glitch I had was systemd spamming /var/log in F20 Alpha TC320:48
@sonney2klisitsyn,  on the other hand we did some releases20:48
adrinbesser82: I know :D ok, gotta go, schones wochenende ;)20:48
besser82adrin:  Dir ebenfalls!  :D20:48
besser82adrin:  Danke20:48
adrinbesser82: systemd is also a hasel in gentoo.20:48
besser82adrin:  in Fedora it is pretty good running now20:49
besser82adrin:  as basically the same ver is targeted for RHEL720:49
-!- ChanServ changed the topic of #shogun to: Shogun Machine Learning Toolbox | Next Stammtisch Monday 2014-01-14 19-24hrs UTC |Devs are usually around between 10am-10pm UTC | When asking questions please be patient or use the mailing list. We always answer but it might take a while. | Channel logs
-!- iglesiasg [] has joined #shogun21:02
-!- mode/#shogun [+o iglesiasg] by ChanServ21:02
@sonney2kiglesiasg, could you please twitter like hell about shogun stuff?21:05
@sonney2kiglesiasg, as in new releases21:05
@iglesiasgsonney2k, all right21:05
@sonney2kiglesiasg, that we want to participate in gsoc 201421:05
@sonney2kand are looking for mentors etc?21:06
@iglesiasgsonney2k, okok21:06
@iglesiasgsonney2k, btw, I think I'd prefer mentoring this year21:06
@sonney2kiglesiasg, and that we are back in debian
@sonney2kiglesiasg, very welcome21:06
@sonney2kiglesiasg, what would you want to do?21:06
@iglesiasgsonney2k, great21:06
@iglesiasgsonney2k, I was thinking of proposing a project in MCMC since it is related to my PhD21:07
@sonney2kmaybe baseline ML algorithms?21:07
@iglesiasgHeiko knows about it too21:07
-!- zxtx_ [] has joined #shogun21:07
@sonney2kI guess HeikoS would want to do it :D21:07
@iglesiasgwe can discuss about it then21:07
@iglesiasgsonney2k, no idea if Heiko this year wants to participate as student?21:09
@iglesiasgI can mentor him lol21:09
@sonney2kiglesiasg, excellent suggestion21:09
@sonney2kiglesiasg, make sure to push him hard21:09
@sonney2kotherwise he will get a side job and learn a new music instrument21:09
@iglesiasghaha yeah21:10
@sonney2kiglesiasg, he just doesn't work w/o pressure ;)21:10
@iglesiasgyour call to retweet now guys21:29
@sonney2kiglesiasg, ok I did all the marketing I could :)21:32
lisitsyniglesiasg: let you mentor me21:32
lisitsynI am still eligible21:32
@iglesiasglisitsyn, haha ok, a pleasure :)21:33
@sonney2klisitsyn, we can play with rose ponies if you'd like21:33
@iglesiasglisitsyn, I would learn from you instead of the other way round though :D21:33
lisitsynsonney2k: I am curious what would be on carol's mind if she see a guy jumping from student to mentor and back21:33
@iglesiasgcurious situation indeed21:35
@iglesiasgbut I'd say it is fine21:35
@iglesiasglisitsyn, would you like to be student this year?21:35
lisitsyniglesiasg: no no21:35
@sonney2klisitsyn, well marry sb. in rossia and you can see the look on her face :D21:35
lisitsynit would be unfair21:35
* sonney2k afk21:35
@iglesiasglisitsyn, I see your point21:36
lisitsynI mean I won't have much time21:36
@iglesiasgbut taking into account some of the projects there have been and the work done...21:36
lisitsynreal students have much more time to enjoy that21:36
lisitsynnot just sponsor some work21:36
@iglesiasgthat is true21:37
lisitsynand I can't predict where I am going to be just in a few months :D21:37
@iglesiasgand well, the idea is also trying to get people that may become contributors21:37
lisitsynsome kind of decision period soon21:37
lisitsynand this sucks phew :D21:38
@iglesiasgthose periods should be exciting :)21:39
* besser82 is off21:40
@iglesiasgbesser82, you could mentor some infra stuff maybe21:40
@wikingiglesiasg: have u managed to check out those things?21:53
@iglesiasgwiking, HR department is having a look and said we would talk this week21:59
@iglesiasgnext week hehe21:59
@sonney2kwiking, what do you want to do?22:17
@wikingiglesiasg: great!22:24
@wikingiglesiasg: fight for your right!22:24
@sonney2knight all22:34
@iglesiasggood nighty22:40
besser82iglesiasg:  infra stuff like?!?23:03
@iglesiasgbesser82, one of the many stuff you usually suggest ;)23:03
@iglesiasgwiking, lisitsyn, anybody there?23:36
-!- iglesiasg [] has quit [Quit: Leaving]23:52
--- Log closed Sat Jan 11 00:00:47 2014