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

Using timezone: Central European Time
* arjanb leaves00:51
* magnus-- leaves01:58
* ChanServ leaves02:07
* Bluelive leaves
* Bluelive joins02:08
* ChanServ joins03:18
* ChanServ leaves03:29
* ChanServ joins04:07
* excerpt joins06:45
* excerpt leaves06:47
* bonniot joins09:55
* Bluelive leaves10:48
* Bluelive joins10:58
* Bluelive leaves11:03
* arjanb joins11:57
* Bluelive joins12:32
<bonniot>it seems i can implement checking for Java method implementations,13:36
and with simple optimizations actually be faster than before :-)13:37
i'm trying to answer brian's mail about return types14:47
<CIA-2>03bonniot * 10Nice/ (5 files in 4 dirs): 15:05
Check that imported Java methods do not have missing implementations.
Be more clever about what Java methods not to test, to save time.
03bonniot * 10Nice/testsuite/compiler/native/implementation.testsuite: 15:24
Check that imported Java methods do not have missing implementations.
Be more clever about what Java methods not to test, to save time.
* magnus-- joins15:29
how are you doing?
arjanb: how is your answer going?15:42
should i make the Any keyword print a deprecated waring in the parser?15:50
<bonniot>yes, OK15:55
arjanb: I think we haven't used yet the full optimizing potential of knowing which parameter positions are not involved in dispatch15:59
i'm looking at that, and it seems like we might win big :-)
didn't get to implementing a change, so it's still suspensefull...16:00
<Bluelive>what do you do at the moment you really think, if only if i had designed this different.. moment16:14
<bonniot>you need to balance the cost to redesign with the benefits16:17
<Bluelive>determining cost is a real problem i think16:19
<bonniot>sure, it's soft knowledge
<CIA-2>03arjanb * 10Nice/ (src/bossa/parser/Parser.jj NEWS): Print deprecated warning when using the 'Any' keyword.
<Bluelive>currently im putting stuff off till the self hosting version16:20
nice allready has depricated features ? :)
<bonniot>sure :-)
<Bluelive>how did that happen ? :)16:21
<arjanb>no much because we remove them after a few months
<bonniot>we're trying to come up to a clean design. you usually don't get everything straight the first time, when you only have incomplete knowledge about the future :-)
yeah. we give some time for users to adapt, still!16:22
<Bluelive>well if you depricate them you still have to support them, instead of just dropping them cold to the floor ?
deprecation is for a transition period
<arjanb>some of the feature are used in the compiler too so taking some time to remove it reduces bootstrap problems
<bonniot>it's usually something like a syntax change16:25
<Bluelive>stupid things like putting private methods in the vtable instead of making them something halfway between an instance and static method16:45
and hidden complexities i hadden accounted for :)16:47
lika A having a private variable O and B having a private variable O
<bonniot>multiple inheritance?16:49
<CIA-2>03arjanb * 10Nice/src/bossa/syntax/Constraint.java: Don't print 'Any' in .nicei16:50
<Bluelive>i decided to just ignore visibility for a while, because i thought it wouldnt have any effect on generated code
which is false16:51
<bonniot>well, it's just an optimization that you are missing, no?16:55
<Bluelive>it can give namecollisions
<bonniot>or if you find field name clashes, but only for non-private fields, then you are not really ignoring visibility ;-)
<CIA-2>03arjanb * 10Nice/stdlib/nice/lang/ (6 files): Removed the use of 'Any' from nice.lang.17:04
<bonniot>arjanb: did you follow Jason's problem with multiple jars on the classpath on windows?17:45
i could reproduce something it on XP on the command line17:46
<arjanb>so it isn't eclipse related?17:53
<bonniot>at least not completely17:55
does it work to specify a drive?
--classpath d:\a.jar;e:\b.jar
for me this didn't work, unless there was "..."17:56
can you reproduce thatM17:58
<arjanb>a moment need to make some jars first17:59
<bonniot>no need to
it's about parsing the command line18:00
the jars don't need to exist
<arjanb>works with "..." but not without18:01
<bonniot>that's a bug. how does it parse it?18:02
it uses both : and ; for splitting?
is " interpreted by the windows shell?18:03
<arjanb>i can't see at the error how it is parsed18:04
<bonniot>yes, you need to debug the console front end
<arjanb>windows treats ; as an space when passing arguments18:12
<bonniot>isn't ; the standard way to specify classpaths on windows?18:13
<arjanb>yes because : are used with drives
<bonniot>so if you write a;b the app gets "a" then "b" in the next arg?18:14
<bonniot>so it's not specific to specifying the drive, is it?18:16
didn't you notice this behaviour before?
<arjanb>i have always written with "..." i didn't know you could that leave away18:17
<bonniot>can you with javac?18:18
maybe it's not windows splitting things at ; but the jvm for windows18:21
D:\Nice>javac -classpath test.jar;test.jar foo.java;bar.java
error: cannot read: foo.java;bar.java
<bonniot>javac is a java program, as far as I know18:22
<arjanb>javac uses a wrapper .exe18:23
<bonniot>can you find out?18:24
<arjanb>how does it handle ; on unixes?18:25
<bonniot>we use : on unix18:28
; is interpreted by the shell as a command separator18:29
<arjanb>i see18:33
<CIA-2>03bonniot * 10Nice/src/gnu/bytecode/ObjectType.java: 18:36
Throw a more characteristic exception when a needed class is missing, so that
we can print an appropriate error message.
<arjanb>not windows but the jvm thinks it needs to interpret the " and ; on the commandline18:37
<bonniot>is that documented somewhere?
<arjanb>i haven't found it yet18:38
not in the jvm spec18:39
*away for meal*18:42
<bonniot>that does not really matter18:43
you should print the args as we get them in main(String[] args) and then see what we can do with them... :-)18:44
D:\Nice>nicec --classpath test.jar;test.jar testbug
<bonniot>that's before calling nice.getopt?18:59
and if you use "" ?
D:\Nice>nicec --classpath "test.jar;test.jar" testbug
<bonniot>it pretty bad, because we cannot distinguish between19:03
nicec --classpath a;b c
(put directories a and b on the classpath, compile package c)
nicec --claspath a b c
(put directory a on the classpath, compile packages b and c)
hum, we could add "" in nicec.bat, no?19:04
(if it's the JVM doing it, we can work around)19:05
<arjanb>i will
hmm maybe it's the batch file interpretting " and ;19:08
it's the batch file19:09
so there is very little we can do except finding a batch file wizard19:15
<bonniot>or some doc? :-)
<CIA-2>03bonniot * 10Nice/src/mlsub/typing/ (TypeConstructor.java Enumeration.java): 19:35
Optimization: really return only one solution for those positions that are not
used for dispatch.
<bonniot>batch files are just brain dead19:50
i guess what we need is a small nicec.exe :-)
<bonniot>we could even write it in nice and compile it with gcj...19:51
<bonniot>do you want to do that?19:59
<arjanb>not now20:05
<bonniot>a simpler approach would be to not reorganize arguments on windows, and to accept successive args to classpath, as long as they are not the last20:27
* ChanServ leaves20:32
* ChanServ joins20:33
* ChanServ leaves20:39
* ChanServ joins21:13