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

Using timezone: Central European Time
<arjanb>have you chosen a different solution to super call problem?00:42
<bonniot>it's commited00:43
<arjanb>yes i have seen that00:44
btw the labeled break changes are committed too00:45
<bonniot>i explored two solutions, but only this one had the good behaviour in all cases00:46
<arjanb>this one looks simple from the outside :-)00:47
<bonniot>simple is beautiful :-)00:49
arjan, please try to include only one aspect in one commit, and not to mix different things00:51
<arjanb>yes i should have commited the parser change separately00:53
do you know a good name for unchecked narrowing conversions?00:57
maybe cast_byte or assume_byte00:58
<bonniot>I think you should not see it as an unsafe feature like this00:59
you can also consider it as "take the byte part (lowest 8 bits) of this integer"01:00
<bonniot>people are used to (byte) i being a truncation and never failing01:01
sometimes it is what you want, sometimes not
so it is useful to have the other way too, but it should not be the default, or some will be surprised (and this can create a runtime exception, so it is serious)
<arjanb>mostly you don't want truncation01:02
<bonniot>so I think we should just add checkByte(...)
who, me? :-)
checkX is good, because it signals that there is a test, and that it might fail01:03
<arjanb>ok i will add these01:04
some ideas of Neel about contructors01:06
but it looks like the development of the Needle language has stopped
<bonniot>he assumes that ? is only useful because of uninitialized fields, which is not true
how do you know?01:07
<arjanb>6 months no activity on mailingslist and cvs
* arjanb leaves01:19
* bonniot leaves
* arjanb joins
* bonniot joins
<arjanb>did you know that's possible to initialize fields before calling the constructor?01:23
it might be helpfull for the initializer in superclass problem01:24
<bonniot>so that syntax bit is still a Q01:35
yeah, I read about that01:36
if I remember correctly, I tried to use this feature, but the bytecode verifier complained01:38
I was probably running JDK 1.3, so it can be that this feature was not enabled until 1.4 (which explains the article, in which he says you need -target 1.4)01:39
this was a bug in previous JDK, apparently, compared to the spec, but that still makes it difficult to use this01:40
(this is from memory, so this would have to be checked again to be sure)
<arjanb>yes so this is not something we could use01:41
for constructors the problem with cyclic dependant classes still isn't solved01:46
<arjanb>class A { B other; } class B { A other; }01:56
or class X { X global; }01:57
now you need to use cast(null) to construct such classes
<bonniot>i see02:00
it's possible to solve by changing the rules for initialized variables02:02
together with visibility: A a = new A(other: new B(other: a));02:06
is there a RFE for this?
<arjanb>this issues comes back regulary so the possible solution should be put on wiki for discussion02:15
should checkedprimitve use assertions or throw something like NumberOverflowException?02:17
anyway time to sleep02:43
<bonniot>could you write the RFE?
did it come back recently?
I only saw asked once (on nice-info I think)
<arjanb>Christian had a similar problem see: http://confer09.condor-edv.com/nice@freenode/2003-10-05.html02:45
second last screen on that page02:46
* arjanb leaves
* bonniot leaves03:49
* arjanb joins10:52

Generated by Sualtam