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

Using timezone: Central European Time
* CIA-4 leaves01:39
* CIA-4 joins
03bonniot * 10Nice/src/bossa/syntax/ (tools.nice MethodDeclaration.java): 01:48
Make sure that existential type variables are confined to the typechecking of
the current method (even if that one has not type variables itself).
<bonniot>good night02:26
<arjanb>good night02:27
* bonniot leaves02:32
* ChanServ leaves06:02
* ChanServ joins06:12
* ChanServ leaves08:34
* ChanServ joins08:52
* bonniot joins11:49
<CIA-4>03bonniot * 10Nice/src/ (2 files in 2 dirs): 11:58
No need to reset the existentialLevel upon exit, it is now guaranteed that
an enter/leave occurs around existential mode.
<arjanb>hmm i don't know all of the conversions that are needed12:53
<bonniot>didn't we say it was just for long[], where there can be Integer and Long?13:02
and in all cases you can use Number.toXxx()
<arjanb>i'm changing SpecialArray now and i don't know what conversions i need in GenericArray so i'm fiddling with bootstrapping everytime13:04
<bonniot>what kind of problem is there?13:05
<arjanb>i need to have the updated GenericArray class in the compiler jar before i can bootstrap13:06
<bonniot>i mean, can you give an example of conversion that is needed?13:08
<arjanb>coercefrom from: ClassType java.util.List13:13
to: Array with unknown element type
coercefrom from: ClassType java.lang.Object
to: Array with unknown element type
coercefrom from: ClassType java.util.Collection
to: SpecialArray(ClassType java.lang.String)
there are quite a few kinds of them
<bonniot>but this is already handled. why do you need to change it?13:14
<arjanb>because GenericArray does a few things differently and represent both array as list and arrays with type params13:16
<bonniot>but why should the conversions be different?13:18
<bonniot>it seems to me you don't need to change any conversion13:23
just to make SpecialArray of unknow elements be compiled as rawArray instead of Object
does that make sense?13:54
why is rawArray.gconvert only used in code generated from SpecialTypes.coerceTo and not in coerceFrom?14:27
<bonniot>convert is used instead?14:33
<arjanb>but gconvert is the only one doing nested arrays
<bonniot>it seems convert_xxx and gconvert_xxx are used when the return type is known, so we also know there are no nested arrays14:36
they are specialized versions
<arjanb>i see14:38
* ChanServ leaves14:39
* ChanServ joins14:40
* ChanServ leaves15:13
* ChanServ joins15:21
* CIA-4 leaves15:38
* CIA-4 joins
<bonniot>about bootstrap: i usually recompile only what has changed, without bootstraping each time, until it works well15:44
only then i adress bootstrap (in the cases where it is not automatic)15:45
that saves a lot of time
can you do that?16:45
<arjanb>not easily
<bonniot>why not?16:46
<arjanb>for testing i need to compile it completely and a batch is not as advanced as a make file16:47
<bonniot>to avoid bootstrap, you only need not to delete already compiled classes16:49
for nice code, nicec will recompile them anyway if necessary
for java code, i usually compile from my editor
but you could also compile *.java16:50
<arjanb>i could try that sometime16:55
<bonniot>arjan, i'm reverting some optimizations on BitVector17:12
with changes to implement type inference, some parts that were not used are needed again17:13
<CIA-4>03bonniot * 10Nice/src/mlsub/typing/lowlevel/BitVector.java:
Revert to 1.4 for now, since later optimizations have changed behaviour
(having a vector longer than the size, it seems), which are problematic now
that we need BitMatrix.transpose again. Need to investigate why.
<bonniot>and that seems to uncover a bug in the optimizations in revisions 1.5 and 1.6
<bonniot>it would be great if you can look at this and find out why
this is a central piece of typechecking, so it's important to make it run fast
it's also important to be sure it's correct too17:15
i will be away during the week-end17:16
<arjanb>when do you leave?
<bonniot>on the other hand, the changes i'm going to make solve one more case :-)
in ~20 minutes17:17
do you have questions?
<arjanb>i expected to have the arrays things ready in an hour or two so i could sent it to you17:19
<bonniot>i will be back sunday afternoon17:20
if it passes all tests and bootstrap, it should be OK :-)17:21
but I can look at it on sunday
if you have time, please look at the BitVector
<arjanb>do you have some code that triggers that bitvector bugs?17:22
<bonniot>yes, i'm commiting it now17:23
basically, BitMatrix.transpose fails with an array index error17:24
normally, a Vector has a size, and all set bits should be smaller than the size17:25
with the recent version of BitVector it seems it was not always the case
<CIA-4>03bonniot * 10Nice/ (2 files in 2 dirs):
Save the constraint when there are any soft variables (which happens when
in existential mode).
<bonniot>if you have the time, it would also be helpful to see what retypings could be removed from nice.lang17:31
<arjanb>finding things to do isn't a problem atm17:32
* CIA-4 leaves
* CIA-4 joins17:33
good luck then17:36
see you on sunday (or monday)
* bonniot leaves17:37
* CIA-4 leaves17:42
* CIA-4 joins
* Bluelive leaves21:02
* kdeuser joins21:06
* kdeuser leaves
* kdeuser joins21:33
can someone help me with sualtam?
can you help any?21:37
<arjanb>i never used sualtam and daniel is away this weekend
<kdeuser>oh ok...
<arjanb>but what is the problem?
<kdeuser>oh... i'm just trying to figure how to use it...
i got it to run and everything
and was wondering when it generates the logs
<kdeuser>i was also wondering if i can change the directory where sualtam saves the logs...
or do i have to have sualtam in the same dir as i want the logs?
<arjanb>i don't see an option for selecting the dir21:42
so i think you have to put sualtam in the same dir
thanks alot
just wondering...
what exactly is this channel for?
<arjanb>discussion about the development of the Nice language (nice.sf.net)21:46
oh i see the topic title is missing
<kdeuser>oh ok...21:47
i was just checking up the online logs that you guys have...
and it's interesting...
uhh...if you would...21:48
thank Daniel for me
i think i'm going to find this little bot real useful21:49
<arjanb>ok :)
take care
see ya
* kdeuser leaves

Generated by Sualtam