Log of the #nice channel on irc.freenode.net

Using timezone: Central European Time
* zzorn_sleep leaves00:43
<arjanb>optimizing the sourceDebugExtension doesn't have much effect01:15
* bangoola_ leaves02:38
* arjanb leaves03:39
* zzorn joins08:27
* bonniot leaves12:17
* bonniot joins13:16
* arjanb joins13:21
<bonniot>i'm writting that part of the dissertation where i'm comparing with multijava atm ;-)16:30
<arjanb>one striking thing of multijava is their absolute requirement of modular typechecking16:50
<bonniot>yes, but i'm showing it's a fake one17:07
we have modular typechecking too
their notion is too restrictive, and because of that forces you to make the compiler happy, even if that mean creating the potential for runtime errors17:08
they recognized it in their paper for relaxed MJ17:09
which allows errors at classloading time
we do these checks purely at compile time
<arjanb>moving checks to the classloader sounds like a bad idea in general17:10
<bonniot>I think their approach would be useful when loading plugins, but it does not seem to be ready for that yet, because you need to pass a list of imported classes when starting the program17:11
<arjanb>i'm somewhat confused about the meaning of 'modular'17:12
<bonniot>that's the crux of the matter17:13
<arjanb>for example is the current implementation of coverage testing in Nice modular?
<bonniot>depends by what you call modular ;-)
<arjanb>redoing coverage testing when adding a class or alternative is not so modular i think17:15
<bonniot>but it's needed
surely there are some more optimizations that could be done, but that's another question
I say that typechecking in Nice is modular17:16
coverage testing isn't, cannot and shouldn't be modular
typically, in MJ, you always need to have a default implementation for methods17:18
so when there is not meaningful default, you can only throw an exception in that case
<arjanb>it would be nice to have a way to avoid value generation always though
<bonniot>then the compiler cannot help you make sure that this default case is never used, and so your program can fail at runtime
value generation? for value dispatch?17:19
<arjanb>i mean the potential exponentional number of tc tuples17:20
<bonniot>i think theoretically it cannot be completely avoided. surely there are more optimisations that could be applied17:22
<arjanb>i guess optimizing that is outside the scope of your dissertation17:24
<bonniot>indeed17:25
and there are already articles about that
what did you find with optimizing the source locations?17:31
<arjanb>that it only saved 10kb on nice.jar and that isn't enough to avoid ArtemGr problem17:33
<bonniot>what percentage is the save? is the optimization complex?17:35
<arjanb>about 15% of SourceDebugExtension attribute17:36
it's using a map to relate file names to a number but i'm not sure if it's correct because i couldn't the format for it17:37
<bonniot>jsr 45?17:39
<arjanb>yes but no specification anywere i looked17:41
<bonniot>you can get the text of the jsr17:42
<arjanb>ah overlooked something17:46
* Atomixx joins17:55
<CIA-1>03arjanb * 10Nice/src/gnu/bytecode/SourceMap.java: Optimization of the SourceDebugExtension attribute size.18:00
* ArtemGr joins18:55
<arjanb>hi19:29
<Atomixx>hi
* ArtemGr leaves19:30
<Atomixx>arjanb, do you know how to build nice.swt.jar?19:34
<arjanb>no but i might be able to help with errors you get19:41
<Atomixx>i download the nice.swt.zip19:44
then i unpack it and run in the directory maven nice:compile
then it tells me it can't download maven-nice-plugin-SNAPSHOT.jar19:45
and the other three packages fail too
<arjanb>i have no idea since i never used maven19:52
<Atomixx>ok no problem19:54
<arjanb>if that jar must come from some sourceforge site that may be the problem19:56
<Atomixx>i thought about that too
but google didn't find the jars19:59
too
<arjanb>http://nice.sourceforge.net/cgi-bin/twiki/view/Doc/MavenGuide
<Atomixx>i've read through it but it doesn't help20:02
unfortunately i must go now
byw
bye
<arjanb>bye20:05
* Atomixx leaves
<bonniot>i'm back, i was away22:39
it seems i always have to leave just when atomixx arrives :-(
pointing to MavenGuide was a good idea22:40
i suspect he might be missing the maven.repo.remote = http://nice.sf.net/maven,http://www.ibiblio.org/maven
<arjanb>i couldn't find that jar anywhere either
<bonniot>it's there
http://nice.sf.net/maven

Generated by Sualtam