<ijp>Stopping for today, I need to figure out how to ensure that cps2 is closure-converted <ijp>ACTION leaves the compile running <catonano>sneek later tell daviid what do you mean with "kawa doesn't impose an eco system or you to compile" ? <catonano>sneek later tell ijp are you working on some cps compilation ? May I ask wat are you doing ? <CustosLimen>so am I misunderstanding boehmgc - or should boehmgc replace free/malloc/realloc ? <CustosLimen>given there is no problem in guile or dependencies itself <cehteh>leaks would be a bug, but if one keeps references to unused objects around then no garbage collector would help ypu <cehteh>glibc allocates some memory once and doesnt free it in some functions, these are non accumulating, shouldnt leak <CustosLimen>cehteh, so as far as I can tell - memory is being leaked but its not memory that is managed by boehmgc - and it seems to be leaking from regex <cehteh>does it accumulate when you add more regex? <CustosLimen>cehteh, it seems to accumlate every time regex is ran <CustosLimen>I did not write the code - and I'm not that familar with guile <CustosLimen>I don't see regfree in guile document - using 2.0.11 <CustosLimen>cehteh, I have a bit of hard time figuring out exactly what code is causing the "leak" that umem perceives <CustosLimen>but not using regfree - not sure how I would even - I see it in code but as far as I can tell its not a lisp function <cehteh>ah i was under the impression that you/guile uses posix regex <cehteh>usually regex are implemented that you run a compiler generating some sort of state machine from the string you give as regex <cehteh>this state machine needs some memory <cehteh>there are plenty other regex implementations and guile may have its own <cehteh>failing to free these state machines would lead to a leak <CustosLimen>cehteh, ok sure - but I dunno how to free them is the point more or less <civodul>it's automatically freed, don't worry :-) <cehteh>could be as simple as dropping all references, then the gc should clean up <cehteh>but you may accidentally keep references around <CustosLimen>civodul, see the problem I'm having is memory just keeps growing <CustosLimen>and I'm seeing that as the source of the leaked memory <CustosLimen>civodul, if it was to be garbage collected it would need to use GC_malloc - but it seems to be just using libc malloc <CustosLimen>civodul, see the reason why I'm woried to beging with is because memory consumption is just growing <civodul>CustosLimen: what tool do you use on this paste? <cehteh>i dunno how guile does that, but it could be that a GC collected object has internally some references to normal allocated objects, and its destructor takes care for a free then <CustosLimen>cehteh, but if that is the case then something is dodgy as umem should have seen references <CustosLimen>anyway I'm just looking for pointers if there are some <civodul>CustosLimen: make-regexp calls regcomp(3) and assumes that regcomp does not allocate anything internally <civodul>it does have a finalizer that calls regfree <civodul>search for '^regexp_free' in libguile/*.c <civodul>CustosLimen: can you (1) come up with a reduced test case that reproduces the problem, and (2) add a printf in regex_free (or breakpoint) to see if it ever gets called <civodul>it could be that regex_free gets called, but not frequently enough <civodul>(because the GC doesn't know that there's a lot of memory associated with the regexp) <civodul>davexunit: with Haunt, should one use (blog #:collections ...) to get articles by tag? <davexunit>I don't know if I have a handy procedure that will just generate a page for each tag <civodul>exactly that, generating a page for each tag <civodul>so that you can then list tags with links <davexunit>I think there's a missing feature here. you can make atom feed pages for each tag, but the blog interface doesn't provide something equally convenient. <davexunit>there is already a posts/group-by-tags procedure <davexunit>it just doesn't fit into the collections thing because collections are declared up front. <davexunit>the tags can't be known until you load the entire posts db <ijp>wingo: I need some help with closure conversion <sneek>ijp, catonano says: are you working on some cps compilation ? May I ask wat are you doing ? <ijp>catonano: I have been trying to convert a guile->js compiler (or most of one) from the previous incarnation of (language cps) to the current one <davexunit>you'd list in in your builders list like (blog-collection-per-tag #:whatever-keyword-args-you-want) <davexunit>I can refine and add something like this to haunt proper sometime <civodul>davexunit: thanks, i'll git it a try! <davexunit>I think I'll start saying that when people ask me for git help <davexunit>civodul: let me know if you are able to make the hack work and I will integrate it into haunt proper sometime, probably instead as a #:tags? flag to the normal blog builder or something. <paroneayea>it seems like if I replace it with just a lambda with the same arguments I get the same result... <paroneayea>I get what stream-cons does, but not stream-lambda <paroneayea>oh, I guess it doesn't evaluate the first one yet. <sneek>wingo, stis says: ping, how goes with the strip functionality vie -g0. Just curious what the state is. <sneek>wingo, stis says: got guile-syntax-parse and guile-log working with guile - looks god, no severe issues. <sneek>wingo, stis says: ping, how goes with the strip functionality vie -g0. Just curious what the state is. <wingo>ACTION has been stuck with de-indirecting structs for a while, been a real mental blocker :P <ijp>wingo: I was just away to hit 'send' on a mail <ijp>well, basically, I've been trying to update my javascript compiler from cps-old to cps-soup. Since I don't have a working master, I've been trying to do it on the two year old branch <ijp>This has been mostly fine as a temporary measure, most of the code is going to translate with no issue, except that I don't have closure conversion <ijp>I tried to see if I could get the current module to work with the old code, but that has not really produced anything <ijp>as a temporary measure, I could reimplement a dump converter, but it seems like a waste to do that <ijp>so basically I was wondering if you could think of any changes that would be preventing me from using the current converter <ijp>as broad and unconcrete a question as that is <ijp>or better, if you have any insight on to the build issue I posted to devel, since then I'd just merge <wingo>regarding the build issue i responded just now ijp <wingo>but unless you are making incompatible changes i would consider staying on stable-2.2 <wingo>dunno, life will probably be easier there <ijp>well, I was having a similar issue on stable-2.2, but I am going to do a make clean before I say anthing else <wingo>not just a make clean, do a "find . '*.go'" just to see <sneek>Welcome back daviid, you have 1 message. <sneek>daviid, catonano says: what do you mean with "kawa doesn't impose an eco system or you to compile" ? <daviid>ACTION is dreaming guile grows up to having large user base and web site like continuum.io ... :) <daviid>sneek: later tell catonano compiling clojure is a nightmare, because you need lenigen or maven (don't recall I abandonned because precisely, you can 'just compile' and I did not want to have to learn thing i don't want to learn ... so in kawa, in a terminal: export CLASSPATH=.:./jars/youjar1.jar:./jars/yourjar2.jar; kawa -S yourkawa.scm just as easy as that, then in other kawa scheme file you can just 'import', like (import (utils)) <daviid>OrangeShark: an 'everything in guile' platform... :) not just science, gui, database, .. but actually we are on a good track, very good track, just a matter of time... <daviid>sneek later tell catonano the compile command is 'kawa -C foo.scm' not kawa -S (I just did hit the wrong key while writing, it is -C you want ...) ***logicmoo is now known as dmles
***dmles is now known as dmioles
***dmioles is now known as dmiles
<daviid>wingo: as it is, does potluck includes what it needs from guix or does it require guix installed 'proper'? <daviid>is it possible to use potluck to globally intall, i.e. in /opt ? <sneek>Welcome back catonano, you have 2 messages. <sneek>catonano, daviid says: compiling clojure is a nightmare, because you need lenigen or maven (don't recall I abandonned because precisely, you can 'just compile' and I did not want to have to learn thing i don't want to learn ... so in kawa, in a terminal: export CLASSPATH=.:./jars/youjar1.jar:./jars/yourjar2.jar; kawa -S yourkawa.scm just as easy as that, then in other kawa scheme file you can just 'import', like (import (utils)) <sneek>catonano, daviid says: the compile command is 'kawa -C foo.scm' not kawa -S (I just did hit the wrong key while writing, it is -C you want ...) <catonano>daviid: I aso share your dream of having an all guile environment <catonano>sneek later tell ijp: ah, a js compiler. Good luck ! <catonano>sneek: later tell daviid thanks. Now I get wat you mean <catonano>sneek: later tell daviid I also share your dream of am all guile environment <dsmith-work>Why is it that every language/platform/environment seems to have this unstoppable urge to reinvent Make ? <cmaloney>because Make works really well for C-line environments and not so great for non-C environments <cmaloney>that and Windows-based installs / compiles <daviid>cmaloney: that is, to me, a false statement, not sure what made you beleive this <sneek>Welcome back daviid, you have 2 messages. <sneek>daviid, catonano says: thanks. Now I get wat you mean <sneek>daviid, catonano says: I also share your dream of am all guile environment <dsmith-work>I think there is some kind of relation to "Greenspun's Tenth" going on. <dsmith-work>I just involved in doing an Android app, and there is yet another builder scripting thingy. <daviid>I'll stick to autoconf/configure/make/texinfo till ... mayyyyybe I give it a go to potluck, to boost guile actually <ijp>wingo: You must have been right about a stale .go, since the build is progressing fine <sneek>Welcome back ijp, you have 1 message. <sneek>ijp, catonano says: ah, a js compiler. Good luck ! <ijp>I'm still a little confused, but I'll make sure and respond on the thread