Open in new window / Try shogun cloud
--- Log opened Sun Nov 24 00:00:38 2013
shogun-buildbot_build #305 of osx1 - libshogun is complete: Failure [failed git]  Build details are at  blamelist: Viktor Gal <>01:29
-!- zxtx [] has quit [Ping timeout: 246 seconds]02:48
-!- zxtx [] has joined #shogun02:58
-!- thoralf [] has joined #shogun03:14
thoralfHello everybody.03:14
-!- zxtx [] has quit [Ping timeout: 252 seconds]03:27
-!- thoralf [] has quit [Quit: Konversation terminated!]03:41
-!- zxtx [] has joined #shogun03:42
shogun-buildbot_build #627 of nightly_default is complete: Success [build successful]  Build details are at
-!- iglesiasg [] has joined #shogun11:24
-!- mode/#shogun [+o iglesiasg] by ChanServ11:24
-!- thoralf [] has joined #shogun12:21
@iglesiasgthoralf, hello hello12:31
@iglesiasgsonney2k, around?12:32
@iglesiasgor someone with some Django experience around?12:32
@iglesiasgok, no problem. I found what I needed :)12:36
thoralfiglesiasg: Trying ony stackoverflow code snippet after another? ;)12:37
@iglesiasghehe not really, I found in the doc actually!12:37
thoralf(don't forget to check the mouse-over of the xkcd strip)12:38
-!- iglesiasg [] has quit [Ping timeout: 245 seconds]13:24
-!- zxtx [] has quit [Ping timeout: 245 seconds]14:28
@sonney2kthoralf, around?15:57
thoralfsonney2k: Yes.16:00
@sonney2kthoralf, I am just wondering why you didn't make SGRefObject not a base class of CSGObject?16:01
@sonney2kbut really only the minimal thing in SGRefObject16:02
thoralfsonney2k: I first wanted some feedback before spending more time on this.16:02
thoralfCSGObject does a lot of things.  Which should be put into SGRefObject and which should stay?  How does this relate to SGReferencedData/ReferencedObject, etc.?16:05
@sonney2kthoralf, just the refcounting no more should go into SGRefObject - as I said my biggest Q would be if we can use SGReferencedData (and just have one base class for referenced stuff)16:06
@sonney2kbut maybe not at this stage16:06
@sonney2kthoralf, then maybe the CDynamicObjectArray should get another parent CDynamicRefObjectArray16:07
@sonney2kwhich contains all the stuff but the serialization16:07
thoralfI knew I'm triggering lots of ideas. ;)16:08
thoralfBut for now, making CSGObject subclass of SGRefObject would do?16:08
@sonney2kthoralf, look the easy (and reasonable change) is SGReferencedObject <- CSGObject16:09
@sonney2kso that should be done in one patch16:09
thoralfYes, I just didn't know if you will approve the way I chose and how much of the functionality should be outsourced.16:31
thoralfIt just happend a few times that I did something you rejected.16:32
@sonney2kthoralf, the absolute minimum16:32
thoralfI'm on the way.16:32
@sonney2kthoralf, you already did too much in one PR...16:32
thoralfFor example?16:33
@sonney2kthoralf, I am not sure how if you should use DynArray later - I mean it does not ref/unref there so it might be leaky16:35
thoralfYes, but it was the easiest way of get it working, since DynamicObjectArray only accepts CSGObjects.16:36
@sonney2kthoralf, it might make sense to have DynamicRefObjectArray for automagic ref/unref'ing16:37
thoralfI totally know that.  But I cannot spend more time on this issue -- the waste of memory has been reduced, so that my stuff runs.16:40
thoralfWhat I now do is just a favour for you guys. ;)16:40
thoralfsonney2k, I'm now having a problem with examples/undocumented/libshogun/library_gc_array.cpp: I'm getting a segfault in "return 0".16:54
thoralfThe problem seems to be that the destructor is called *after* shogun_exit().16:54
thoralfPutting a scope around, ending before shogun_exit() solves the problem.16:56
thoralfObviously we shouldn't call shogun_exit() if we're having automatic (shogun) variables in the same scope.16:58
* sonney2k is just doing this
thoralfThis video is currently unavailable.17:05
thoralfBut the title tells a lot.17:05
@sonney2kthoralf, the simple refactoring shouldn't break anything17:13
thoralfsonney2k: Yes, but the problem existed before.17:14
thoralfcalling shogun_exit() in a scope, where variables exist, is problematic: In my case only changing SG_GCDEBUG to SG_SGCDEBUG triggered the problem.17:16
@sonney2kthoralf, yes of course makes sense17:22
@sonney2keverything has to be freed before that17:23
-!- zxtx [] has joined #shogun17:27
thoralfsonney2k: The pull request has been updated.  I'm happy to announce: All refcount code from CSGObjects has been put into its new superclass SGRefObject.17:35
@sonney2kthoralf, errm dont' see the PR!17:42
thoralfIt's still this one:
thoralfpush --force17:42
thoralfI did not create a new class DynamicRefObjectArray, btw.17:44
@sonney2kthoralf, ohh please3 a PR with just the SGRefObject no more17:46
thoralfWhat about the StructuredData stuff?17:46
@sonney2kthoralf, not yet17:48
thoralfI don't understand.17:50
-!- thoralf [] has quit [Quit: Konversation terminated!]18:19
@wikinglisitsyn: ping18:46
-!- sonne|osx [] has joined #shogun19:23
sonne|osxargh thorsten left19:24
sonne|osxhey wiking19:24
lisitsynwiking: pong19:33
lisitsynwiking: what kind of morning it is? :D19:34
lisitsynare you in the us or what? ;)19:34
-!- thoralf [] has joined #shogun19:38
-!- sonne|osx [] has quit [Quit: sonne|osx]19:42
-!- sonne|osx [] has joined #shogun20:52
-!- lisitsyn [~lisitsyn@] has quit [Ping timeout: 264 seconds]20:53
-!- lisitsyn [~lisitsyn@] has joined #shogun20:54
sonne|osxthoralf: what I meant is that we should for now just do the SGRefObject refactoring in one PR not the other stuff you are doing21:11
sonne|osxthoralf: and really minimal please - no other functions like get_name etc - these should really all stay in SGObject21:11
sonne|osx(to the extent possible...)21:11
thoralfsonne|osx: No problem, I'll send another PR.  But remember how it started: You said we need this StructuredData refactoring.21:14
thoralfThat's why I'm confused.21:14
sonne|osxthoralf: yeah but that was before you realized that you need refcounting21:15
thoralfsonne|osx: Btw. I found method stubs deep_copy() and a shallow_copy() in CSGObject.21:20
thoralfThey are not used.21:20
thoralfRemove them?21:20
sonne|osxthoralf: not used at all? then remove them.21:21
thoralfThe stubs are SG_NOTIMPLEMENTED, not nowhere overridden.21:21
thoralfSG_SGCDEBUG("unref() refcount %ld obj %s (%p) decreased\n", count, this->get_name(), this)21:26
thoralfget_name() is required in SGRefObject ;)21:26
sonne|osxthoralf: that was only for debugging :/21:43
sonne|osxbut hmmh don't see any better option21:43
thoralfNo, I found shallow_copy() implemented in RelaxedTree.21:44
thoralfand GaussianKernel21:44
sonne|osxthoralf: but this can be in CSGObject21:46
-!- sonne|osx [] has quit [Quit: sonne|osx]22:01
-!- thoralf [] has quit [Quit: Konversation terminated!]23:11
-!- thoralf [] has joined #shogun23:25
-!- pickle27 [] has joined #shogun23:39
-!- pickle27 [] has left #shogun []23:39
--- Log closed Mon Nov 25 00:00:39 2013