2017-06-14T02:31:13Z kini quit (Quit: No Ping reply in 180 seconds.)
2017-06-14T02:33:40Z kini joined #ccl
2017-06-14T03:10:10Z rumbler31 quit (Remote host closed the connection)
2017-06-14T03:35:34Z rme: dim: I have two suggestions.  One is to run ccl with the ''--batch" option.  This will print out some information when the lisp crashes.  It may provide a clue.
2017-06-14T03:37:01Z rme:  The second thing might be to try running ccl with "--heap-reserve 150g".  Perhaps the container environment doesn't like ccl's default reservation of 512 gigabytes of address space.
2017-06-14T03:37:46Z rme: Note that it there's a lower limit of about 128g, so trying to reserve less than that doesn't have any effect.
2017-06-14T03:50:01Z rumbler31 joined #ccl
2017-06-14T03:54:40Z rumbler31 quit (Ping timeout: 246 seconds)
2017-06-14T04:28:51Z dustinm` quit (Ping timeout: 260 seconds)
2017-06-14T04:29:14Z dustinm` joined #ccl
2017-06-14T04:30:04Z rumbler31 joined #ccl
2017-06-14T04:34:10Z rumbler31 quit (Ping timeout: 240 seconds)
2017-06-14T05:16:26Z rumbler31 joined #ccl
2017-06-14T05:20:46Z rumbler31 quit (Ping timeout: 246 seconds)
2017-06-14T05:30:48Z rumbler31 joined #ccl
2017-06-14T05:36:52Z rumbler31 quit (Ping timeout: 246 seconds)
2017-06-14T06:00:30Z Intensity quit (Remote host closed the connection)
2017-06-14T06:03:44Z mdc_mobile quit (Ping timeout: 255 seconds)
2017-06-14T06:26:16Z Intensity joined #ccl
2017-06-14T06:33:21Z rumbler31 joined #ccl
2017-06-14T06:37:32Z rumbler31 quit (Ping timeout: 246 seconds)
2017-06-14T06:54:10Z sabetts quit (Ping timeout: 240 seconds)
2017-06-14T08:12:47Z hhdave joined #ccl
2017-06-14T08:56:11Z dim: will try both, thanks for the hints!
2017-06-14T08:56:40Z dim: is heap reserve comparable to “overcommit” memory behavior of the JVM?
2017-06-14T09:06:32Z rumbler31 joined #ccl
2017-06-14T09:10:50Z rumbler31 quit (Ping timeout: 246 seconds)
2017-06-14T09:23:16Z fiddlerwoaroof quit (Ping timeout: 268 seconds)
2017-06-14T09:24:37Z fiddlerwoaroof joined #ccl
2017-06-14T09:29:44Z mrottenkolber joined #ccl
2017-06-14T09:47:13Z dim: (as I'm using buildapp, https://github.com/xach/buildapp/issues/31)
2017-06-14T10:19:48Z fiddlerwoaroof quit (Ping timeout: 260 seconds)
2017-06-14T10:27:25Z fiddlerwoaroof joined #ccl
2017-06-14T10:49:26Z dim: I added the --batch option but it's not failing at a stage where this option has any impact: https://hub.docker.com/r/dimitri/pgloader/builds/b8ahc5hmrjqwneskude5iht/ and https://travis-ci.org/dimitri/pgloader/jobs/242761913
2017-06-14T10:50:58Z fiddlerwoaroof quit (Ping timeout: 260 seconds)
2017-06-14T10:59:28Z rumbler31 joined #ccl
2017-06-14T11:03:57Z rumbler31 quit (Ping timeout: 240 seconds)
2017-06-14T11:06:43Z pjb quit (Ping timeout: 255 seconds)
2017-06-14T11:39:19Z fiddlerwoaroof joined #ccl
2017-06-14T11:49:35Z fiddlerwoaroof quit (Ping timeout: 240 seconds)
2017-06-14T12:07:05Z pjb joined #ccl
2017-06-14T12:22:46Z pjb quit (Ping timeout: 255 seconds)
2017-06-14T12:29:07Z rumbler31 joined #ccl
2017-06-14T12:44:10Z mrottenkolber quit (Ping timeout: 240 seconds)
2017-06-14T13:09:01Z fiddlerwoaroof joined #ccl
2017-06-14T13:16:46Z fiddlerwoaroof quit (Ping timeout: 246 seconds)
2017-06-14T13:21:57Z rumbler31 quit (Remote host closed the connection)
2017-06-14T14:22:31Z rumbler31 joined #ccl
2017-06-14T14:24:24Z fiddlerwoaroof joined #ccl
2017-06-14T14:27:36Z rumbler31 quit (Ping timeout: 260 seconds)
2017-06-14T14:31:47Z fiddlerwoaroof quit (Ping timeout: 255 seconds)
2017-06-14T14:34:03Z rumbler31 joined #ccl
2017-06-14T14:45:27Z fiddlerwoaroof joined #ccl
2017-06-14T15:11:57Z fiddlerwoaroof quit (Ping timeout: 240 seconds)
2017-06-14T15:23:36Z rumbler3_ joined #ccl
2017-06-14T15:27:57Z rumbler3_ quit (Ping timeout: 240 seconds)
2017-06-14T16:23:32Z sabetts joined #ccl
2017-06-14T16:24:40Z rumbler3_ joined #ccl
2017-06-14T16:29:50Z rumbler3_ quit (Ping timeout: 240 seconds)
2017-06-14T16:38:33Z pjb joined #ccl
2017-06-14T16:48:48Z hhdave quit (Ping timeout: 240 seconds)
2017-06-14T17:35:50Z xrme joined #ccl
2017-06-14T17:37:34Z xrme: dim: I don't exactly how you are making the "pgloader" executable.  If it's a shell script that execs ccl somewhere, add the "--batch" flag there.  If it's a saved image with a prepended lisp kernel, then it should work to pass "--batch" to pgloader.
2017-06-14T17:38:43Z xrme: In other words, I think you want to say "./regress.sh /home/travis/build/dimitri/pgloader/build/bin/pgloader --batch csv-keep-extra-blanks.load"
2017-06-14T17:40:28Z xrme: You might also try ../pgloader --heap-reserve 130g --batch
2017-06-14T17:42:17Z xrme: If we find out that ccl's technique of reserving a large amount of virtual address space is causing problems in container-based deployments, that will require some reflection on whether we should do something about it.
2017-06-14T17:44:26Z xrme: https://github.com/Clozure/ccl/issues/25 is another issue where reserving a lot of address space is causing some irritation.
2017-06-14T17:47:53Z mrottenkolber joined #ccl
2017-06-14T17:52:07Z xrme: dim: It's a pity that I'm 9 hours behind you, timezone-wise...
2017-06-14T17:53:53Z xrme: mrottenkolber: Hi. I look forward to more ccl commits from you!
2017-06-14T17:54:50Z jdz quit (Ping timeout: 240 seconds)
2017-06-14T18:01:38Z jdz joined #ccl
2017-06-14T18:01:50Z mrottenkolber: xrme: Hey :-) I look forward to contribute more. My next goal is to polish ccl:conditional-store (extend the types of places it accepts, make it work with with-slots).
2017-06-14T18:41:22Z dim: hi
2017-06-14T18:42:35Z xrme: Hi dim
2017-06-14T18:42:53Z dim: xrme: yeah well TZ... we'll work around it, and also I don't want to steal so much of your time, so as long as pgloader is triggering a ccl bug and we have a common interest that's cool, but if that's not a priority for you then let's do differently
2017-06-14T18:44:05Z dim: currently I build the image with buildapp from Xach, it's a cool tool and supports both SBCL and CCL ; and I parse command line arguments with command-line-arguments from qitab's org (now Google IIUC)
2017-06-14T18:45:13Z xrme: There are certain command-line arguments that CCL's runtime will always intercept and process.  Others it will pass along to lisp.
2017-06-14T18:46:49Z xrme: "--batch" and "--heap-reserve" are examples of that
2017-06-14T18:48:59Z xrme: Some people don't like this.  https://trac.clozure.com/ccl/ticket/605 and https://trac.clozure.com/ccl/ticket/864
2017-06-14T18:54:36Z mrottenkolber: You can avoid this without a shell wrapper
2017-06-14T18:55:23Z mrottenkolber: https://github.com/eugeneia/erlangen/blob/master/build/port-mapper.lisp#L35
2017-06-14T18:56:00Z mrottenkolber: TL;DR You need to define an application-class
2017-06-14T18:56:41Z mrottenkolber: https://github.com/eugeneia/erlangen/blob/master/build/port-mapper.lisp#L22
2017-06-14T18:57:28Z mrottenkolber: If I remember correctly I figured that out by reading the source, so it might be that this solution is less supported than I think it is
2017-06-14T18:57:47Z mrottenkolber: xrme can probably confirm if that’s a good thing to do
2017-06-14T19:01:38Z xrme: I seem to recall that there's no way for your application to process, say, "-n", because the lisp kernel always processes that.  Maybe I am not remembering correctly.  I would need to check.
2017-06-14T19:04:01Z xrme: When specifying a heap image, you now have to use "-I heap-image.image", so the example from Trac ticket 605 doesn't fail as it once did.
2017-06-14T19:04:11Z xrme: I am certainly babbling away today.
2017-06-14T19:06:05Z mrottenkolber quit (Ping timeout: 240 seconds)
2017-06-14T19:15:10Z dim: pushed --batch and --heap-reserve 150g in the test env, but I can't build locally with CCL currently because of missing libs (I need to link with some .so) --- need to fix that too I guess
2017-06-14T19:16:29Z xrme: Here's hoping that will provide a clue about what is going wrong.
2017-06-14T19:32:08Z dim: ok as soon as I start working correctly (3rd useless commit/push to travis) we might have a chance
2017-06-14T19:38:57Z dim: ok, there you go: https://travis-ci.org/dimitri/pgloader/jobs/242972894#L1370
2017-06-14T19:39:02Z dim: Unrecoverable stack overflow.
2017-06-14T19:39:12Z dim: on the first test case, which used to work
2017-06-14T19:43:08Z xrme: I think that's probably not the same problem as before;  it is probably a side-effect of "--batch".
2017-06-14T19:46:05Z xrme: Perhaps it is best to try to duplicate the problem locally before trying anything else in the test environments.
2017-06-14T20:11:14Z dim: the fact that if fails on a test-case that used not to makes me think the same
2017-06-14T21:32:40Z mrottenkolber joined #ccl
2017-06-14T21:33:09Z mrottenkolber: xrme: no, it seems to even disable -n
2017-06-14T21:34:01Z mrottenkolber: i.e. in my example -n is passed as the first argument (which fails in this specific case bacause its not a valid hostname)
2017-06-14T21:35:23Z rumbler3_ joined #ccl
2017-06-14T21:38:08Z pierpa joined #ccl
2017-06-14T21:38:17Z rumbler3_ quit (Remote host closed the connection)
2017-06-14T21:45:32Z mrottenkolber: xrme: I just surfed through github, and stumbled on thoughstuff.com/rme, so you work with Clojure currently? How do you like it? My first gig was actually a Clojure program, that modeled hierarchical neural nets. The project was hence super interesting, but I got mixed feelings with Clojure (Lisp yay, Java stack trace nay).
2017-06-14T21:50:35Z mrottenkolber: Oh, and you can fly a plane. Good to know. ;-)
2017-06-14T21:53:51Z rumbler3_ joined #ccl
2017-06-14T21:58:37Z rumbler3_ quit (Ping timeout: 246 seconds)
2017-06-14T22:23:07Z rumbler31 quit (Ping timeout: 246 seconds)
2017-06-14T22:54:54Z rumbler31 joined #ccl
2017-06-14T23:00:10Z rumbler31 quit (Ping timeout: 240 seconds)
2017-06-14T23:01:04Z rme: mrottenkolber: I like CL better.
2017-06-14T23:07:42Z rme: On the one hand, the ability to call Java code easily seems like a big benefit.  On the other hand, I find that it's pretty common for Java to show through a lot.
2017-06-14T23:08:21Z rme: And like you say, stack traces and debugging could be better.
2017-06-14T23:10:04Z rme: My feeling also is that Clojure goes too far in removing "excess" parens.  This makes cond forms harder to use, IMO.  And C-M-t doesn't work to re-order bindings in let forms.
2017-06-14T23:10:45Z rme: I also don't care for the use of square brackets in addition to parens.  And why is it "defn" instead of "defun"?  Does Clojure have something against fun?  :-)
2017-06-14T23:13:33Z rme: But maybe I'm just so used to CL's idiosyncrasies that I have grown to like them.
2017-06-14T23:24:54Z rme: And finally (ha), I'm coming to Clojure from a traditional Lisp background.  I think a large number of people come to it from Java. I think that experience affects how they use they use the language.  It affects the developement workflow they choose to use, too.  I kind of take repl-driven development as the standard way to do things, but to my surprise, not everyone works that way.
2017-06-14T23:26:52Z rumbler31 joined #ccl
2017-06-14T23:31:28Z rumbler31 quit (Ping timeout: 255 seconds)
2017-06-14T23:32:49Z rumbler31 joined #ccl
2017-06-14T23:34:12Z rumbler31 quit (Remote host closed the connection)
2017-06-14T23:43:22Z mrottenkolber: xrme: agree with all your points, I kinda lost it when I stumbled upon with-out-str, a butchered with-output-to-string that always captures standard output
2017-06-14T23:44:37Z mrottenkolber: I feel like ABCL can do Java interop pretty well, plus nowadays its a pretty complete implementation with good support
2017-06-14T23:51:12Z mrottenkolber: When I asked my then client why he specifically wanted clojure and not say CL, he somewhat ridiculed me saying that he doesn’t want to invest in an obscure language. I thought that was really unfair, I considered Clojure arguably even more obscure back then, and still do. I suppose its easier to hire Clojure devs, but I also take into account that CL has a proven language standard and many implementations while Clo
2017-06-14T23:51:12Z mrottenkolber: ion language, with no specification. Back then the documentation of the core language was also lacking.
2017-06-14T23:53:20Z rme: I think "here, take this .jar file" is a pretty attractive thing to be able to say in certain organizations.
2017-06-14T23:55:01Z mrottenkolber: New theory: Clojure is designed specifically to appeal to devs that are used to worse-is-better. I.e. (with-output-to-string (stream) ...) vs (with-out-str ...). Maybe, it would raise eyebrows in the non-CL world if names were not noxiosly abbreviated ;-)
2017-06-14T23:55:46Z mrottenkolber: True, I think ABCL can do that. I recall Mark Evenson talking about deploying ABCL generated JARs with Tomcat.
2017-06-14T23:56:00Z rme: On the other hand, I'm not sure how that's all that much better than "here, take this single binary, which is a saved image with a prepended lisp kernel".
2017-06-14T23:56:34Z mrottenkolber: For the companies its specifically important to be a JAR, because their ops knows how to deal with JARs
2017-06-14T23:56:58Z mrottenkolber: I.e. there is a fixed requirement “has to be a JAR/WAR that goes into Tomcat”
2017-06-14T23:57:06Z rme: That could be.
2017-06-14T23:57:15Z mrottenkolber: Java is BIG
2017-06-14T23:58:05Z rme: There are many good ideas in Clojure.  When I've had the chance to listen and talk to Rich Hickey, it's clear that he's given serioius thought to the language and the abstractions it provides, and the things it makes easy.
2017-06-14T23:58:43Z rme: So I don't want to come across as someone who is just slagging off Clojure.  That's not true.