Open in new window / Try shogun cloud
--- Log opened Thu Aug 04 00:00:51 2016
-!- OXPHOS [4441a916@gateway/web/freenode/ip.] has joined #shogun01:50
@wikingOXPHOS, ping01:54
OXPHOSwiking: hey sgvector serialization should be working now. I just squashed01:54
@wikingyeah i know01:54
@wikingbut i just replied :)01:55
@wikingread it and if you are open to talk a bit01:55
@wikinglets talk :)01:55
OXPHOSwiking: kk01:55
OXPHOSwiking: also I had problem fetching from shogun. I thought it was the network but seems not..01:55
@wikingfetching what?01:55
OXPHOSMac148934:shogun zora$ git fetch upstream01:55
OXPHOSfatal: unable to connect to[0:]: errno=Operation timed out01:55
OXPHOSwiking ^01:56
@wikingcheck status.github.com01:56
@wiking``` All systems operational ```01:56
@wikingbut who knows...01:56
OXPHOSwiking: I can fetch shogun-data01:56
OXPHOSwiking: and I was experiencing the same problem ~12h ago..01:56
@wikingping github on twitter01:57
@wikingi bet they'll answer in no time :)01:57
OXPHOSwiking: so it's their problem : ) lemme try01:57
@wikingyep i mean all our code is on github01:58
@wikingso yeah i cannot help01:58
OXPHOSwiking: kk thx01:58
-!- shogun-notifier- [] has joined #shogun02:07
shogun-notifier-shogun: christopher :develop * ade10a6 / / (4 files):
shogun-notifier-shogun: LARS float32 / floatmax support clean up02:07
shogun-notifier-shogun: Replacing if-SGERROR sets with REQUIRE statements02:07
shogun-notifier-shogun: Fixing memory leaks / implicit type conversion warnings02:07
shogun-notifier-shogun: Viktor Gal :develop * ed1c24e / / (4 files):
shogun-notifier-shogun: Merge pull request #3401 from c4goldsw/develop02:07
shogun-notifier-shogun: LARS float support cleanup patch (for PR #3133)02:07
shogun-buildbotbuild #1319 of precise - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Viktor Gal <>, christopher <>02:21
shogun-buildbotbuild #2961 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Viktor Gal <>, christopher <>02:30
shogun-buildbotbuild #98 of xenial - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Viktor Gal <>, christopher <>02:31
@wikingso the problem is we just cannot include cereal headers02:46
@wikingunless we make cereal dependency a STRONG dependency02:46
OXPHOSwiking: like Eigen?02:49
@wikingwell no not really02:49
@wikingbecause in eigen the problem is solved internally02:49
@wikingmeaning that although we download eigen for building shogun02:49
@wikingbut we dont distribute it02:49
@wikingbecause we compile the header related stuff into shogun02:50
@wikingbut what you are doing now is actually making the cereal dependency outside of shogun02:50
@wikingimagine that you compile and install libshogun02:50
@wiking(without the cereal headers)02:50
@wikingi mean without having cereal on the system02:50
@wikingthen you'll download cereal to build libshogun02:51
@wikingand then ...?02:51
@wikingif the user creates a new example02:51
@wikingin c++ where he uses shogun02:51
@wikinghe will not be able to compile it at all02:51
@wikingbecause a) he doesn't have cereal installed on the system02:51
@wikingb) the headers you used for shogun to compile are long gone02:52
@wikingdo you get what i'm trying to say here?02:52
@wikingin eigen case things are all wrapped into .cpp02:52
OXPHOSwiking: not they have to be moved away from .h files in shogun?02:54
@wikingso yeah02:55
@wikingbasically the idea is02:55
@wikingthat if we cannot solve02:55
@wikingnot to include cereal headers02:55
@wikingin our headers02:55
@wikingthen we have to make cereal a strong dependency02:55
@wikingnote: not having some type defined is fine, because those can be forward declared in headers02:56
@wikingsee the trick about Eigen in SGVector02:56
OXPHOSwiking: you mean the namespace Eigen{} stuff?02:57
@wikingthat is all because02:58
@wikingwe do not want to include eigen related headers in our headers02:58
@wikingonly in cpp02:58
OXPHOSwiking: not checked closely yet - but there's include <Eigen/xx> in mathematics/eigen3.h02:59
@wikingand where do we include eigen3.h? ?:)02:59
@wikingi mean trust me03:00
@wikingi know what i'm talking about ehre03:00
@wikingno need to check things like this03:00
OXPHOSwiking: yeah i believe you are :) ..trying to make myself clear. so the goal is to move cereal stuff to .cpp03:01
@wiking(btw you do realise that mathematics/eigen3.h is just a wrapper around Eigen/xxxx headers )03:01
@wikingbut yeah the idea is that03:01
@wikingin headers03:01
@wikingthat exposed to the public03:01
@wiking(say not linalg related headers)03:01
@wikingyou should not include cereal headrs03:02
@wikingonly in definition03:02
@wikingmeaning cpp03:02
@wikingif that's not solveable03:03
@wikingwe'll have to default to the idea03:03
@wikingthat we gonna make cereal a strong dependency03:03
@wikingbut let's try to concentrate now on the03:03
OXPHOSokay..I think I got it03:03
@wikingpart where we could solve this03:03
@wikingand btw yeah it all makes sense03:03
@wikingwhy it was working for you on your machine03:03
@wikingas i believe you have cereal installed on your system03:04
@wikingwhile on travis it was using the internal cereal (fetching as the part of cmake build...)03:04
OXPHOSI tried removed it - and it also worked..I guess some settings are different03:04
@wikingit should not03:04
@wikingbut ok03:04
@wikingthe problem is this definitely atm03:05
@wikingthat we act like we have cereal headers installed globally03:05
@wikingwhere we fake it sometimes...03:05
OXPHOSlemme work on it03:06
@wikingnw thnx for looking at it03:06
@wikingcan you do rebases on some of the prs?03:06
@wikingor github is not working for you atm?03:07
OXPHOSwiking: not working - only fetch upstream is not working03:07
@wikingcan you copy paste the exact error lines?03:07
OXPHOSwiking: 1 sec. it takes a while to know it's failed :)03:09
@wikingcan you03:10
@wikingping github.com03:10
OXPHOSwiking: PING ( 56 data bytes 64 bytes from icmp_seq=0 ttl=52 time=68.613 ms03:12
OXPHOSand so on..03:12
@wikingmmm yeah03:12
@wikingok try03:13
@wikingok not this03:13
@wikingnc -vz 44303:14
@wikingand then03:14
@wikingnc -vz 44303:14
OXPHOSwiking: succeeded03:15
@wikingcat shogun/.git/config03:16
OXPHOSwiking: I should do it in dir upstream shogun right?03:18
OXPHOSMac148934:Github zora$ cd shogun/.git/config -bash: cd: shogun/.git/config: Not a directory03:18
@wikingi mean within your shogun just do03:18
@wikingcat .git/config03:18
-!- OXPHOS_ [4441a916@gateway/web/freenode/ip.] has joined #shogun03:21
-!- OXPHOS [4441a916@gateway/web/freenode/ip.] has quit [Ping timeout: 250 seconds]03:22
OXPHOS_wiking: sry my irc somehow got stuck..03:22
@wikingwhen have you added the upstream to your config?03:23
@wikingwas it ever working?03:24
@wikingas i believe you dont have access rights to the upstream repo03:24
@wiking[remote "upstream"]03:24
@wiking    url = git://
@wikingand this would only work in that case03:24
@wikingyou should edit your config03:24
@wikingand use https03:24
@wikingurl =
@wikingand yeah that git thign was toootaly weird03:25
OXPHOS_what..I never touched it can I used it pretty frequently..03:25
@wikingbecause if then it should have been03:25
@wikingso i dont really know how this got into this state03:25
@wikingbut no wonder it's not working03:25
OXPHOS_wiking: haha thanks! it must be hacked..or maybe wired : /03:26
OXPHOS_wiking: it's back to work!03:28
@wikingso rebase now plz03:29
OXPHOS_wiking: i'm running(checking) it at local now. I'll push once it's done.03:38
shogun-buildbotbuild #1181 of nightly_default is complete: Failure [failed test notebooks]  Build details are at  blamelist: Viktor Gal <>, christopher <>03:57
-!- OXPHOS_ [4441a916@gateway/web/freenode/ip.] has quit [Ping timeout: 250 seconds]04:24
-!- OXPHOS [4441a916@gateway/web/freenode/ip.] has joined #shogun05:03
-!- shogun-notifier- [] has quit [Quit: transmission timeout]05:07
shogun-buildbotbuild #47 of FC23 - libshogun - aarch64 is complete: Failure [failed test]  Build details are at  blamelist: Viktor Gal <>, christopher <>05:36
shogun-buildbotbuild #53 of memleak - valgrind is complete: Failure [failed memory check]  Build details are at  blamelist: Viktor Gal <>, christopher <>06:37
shogun-buildbotbuild #1052 of nightly_none is complete: Failure [failed compile]  Build details are at  blamelist: Viktor Gal <>, christopher <>06:40
-!- OXPHOS [4441a916@gateway/web/freenode/ip.] has quit [Ping timeout: 250 seconds]08:32
-!- leagoetz [] has joined #shogun10:06
-!- leagoetz [] has quit [Remote host closed the connection]10:17
@wikingSaurabh7, ping10:41
-!- sanuj [~sanuj@] has joined #shogun10:54
sanujlisitsyn, wiking yo\10:54
sanujlisitsyn, are you working on SG_ADD?10:54
lisitsynsanuj: yeah will spend some time on that today10:54
lisitsyndidn't get time yesterday tho10:55
sanujlisitsyn, okay, how to make swig interfaces for shogun templates like SGVector?10:55
lisitsynnot get it10:55
lisitsynwe specialize sgvector for specific types10:56
sanujlisitsyn, shogun-base.i looks like this:
lisitsynonce specialized it is just some time10:56
sanujwe had left the templates since you wanted to discuss with HekioS and wiking10:56
sanujlisitsyn, so i just do SGVector<blah>10:56
lisitsynyeah I think so10:56
sanujeven i think there is no way10:57
sanujlisitsyn, so i'll just do SGVector with some primitive types, just for demonstration10:57
@wikingis SGVector<float64_t>10:59
@wikingi dont really understand the swig interface for shogun templates like sgvector Saurabh710:59
sanujwiking, yes yes11:00
@wikingmeaning? :)11:16
@wikingwe dont use SGVector for other types than primitive types11:17
@wikingfor none primitive types there are other data structs11:17
sanujwiking, i mean i got it now :)11:17
sanujhad forgotten about RealVector etc11:18
-!- sanuj [~sanuj@] has quit [Ping timeout: 252 seconds]13:52
-!- sanuj [~sanuj@] has joined #shogun13:53
sanujlisitsyn, there?15:36
lisitsynsanuj: yeap15:36
-!- sanuj [~sanuj@] has quit [Ping timeout: 252 seconds]15:40
@wikingthat was fast15:42
-!- sanuj [~sanuj@] has joined #shogun17:19
sanujwiking, yo17:48
@wikingwhat's happening17:48
sanujwiking, can you review this:
sanuji think it's done17:48
sanujwiking, MockObject is also used for integration testing17:58
@wikingdont care17:59
@wikingif it's for testing17:59
@wikingthen it's no place17:59
@wikingfor that in code created for runtime17:59
sanujwiking, so where do i put it then?18:00
@wikingfor tests18:00
@wikingwhere ever they need to be18:00
@wikingbut not into a runtime18:00
@wikingwhere you possibly not going to be using it18:00
@wikingif you think there's no way you can solve it18:01
@wikingtests for macros are still possible18:01
@wikingalthough again i dont understand that why does a mock  object18:01
@wikinghas to be in runtime18:01
sanujwiking, i am not clear what you mean by runtime?18:01
@wiking src/interfaces/modular/SGBase.i18:02
sanujand MockObject is in SGBase.i so that it is available in python for integration tests18:02
@wikingis something that will be created18:02
@wikingno matter18:02
@wikingif you do ever an integration test18:02
@wikingor not18:02
@wikingi mean  src/interfaces/modular/SGBase.i  -> modshogun.i -> modshogun<interface>_wrap.cxx18:02
@wiking-> swig SO library18:03
sanujwiking, it has to be in modshogun.i, one way or another18:03
@wikingok so if i'm user HailMerry18:03
@wikingwho just wanna use shogun18:03
@wikingfrom python18:03
@wikingdo i need that?18:03
sanujwiking, so only make it available when enable_testing=on?18:04
@wikingnow you understand?18:14
sanujwiking, yeah, so i only need to add some MACRO checks?18:14
sanujjust confirming18:14
@wikingidk if there's actually a C level MACRO when we are running tests :)18:15
sanujwiking, TagRealVector and RealVector are different because TagX is used to identify an object of X by a name18:28
sanujshall i put this in modshogun.i?18:30
sanuj+#ifdef HAVE_JINJA218:30
sanuj +%include "shogun-base.i"18:30
sanuj +#endif18:30
sanujinstead of SGBase.i18:30
sanujwiking, by doing this, i can write this as18:30
sanujRealVector, RealVector*18:31
-!- sanuj [~sanuj@] has quit [Remote host closed the connection]20:36
--- Log closed Fri Aug 05 00:00:52 2016