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

Using timezone: Central European Time
* bonniot leaves00:55
* arjanb leaves01:56
* zzorn joins07:50
* ArtemGr joins10:27
<CIA-1>03fbarber * 10Nice/src/nice/tools/doc/main.nice: Move compilation variable to be a global var (allows to work out if we are compiling a given package later on).11:41
* bonniot joins11:51
<ArtemGr>I've recompiled FreeBSD JDK with debugging support (can use gdb), but my program currently aborts with some hotspot cpp assertion when started in the debug-enabled java, so no luck here yet.12:33
<bonniot>is this about the JVM internal error?12:34
<bonniot>did you try a bytecode verifier?12:37
<ArtemGr>Obfuscator is not involved, it is just an -Xdebug error, I'm pretty sure I can repeat the error with no obfuscator runs at all.12:40
<bonniot>and what does -Xdebug do?12:42
<ArtemGr>Where you present yesterday in IRC? See the log if not.12:43
<bonniot>i was. but i still don't know precisely what -Xdebug is good for12:47
<ArtemGr>It allows remote debugging, used with jdb, JSwat, whatever.
could you manage to point out the bad symbol?12:49
<ArtemGr>I was going to use the gdb for it. I have access to the source where the exception (InternalError) is thrown, but the code is foreign to me and i don't know a safe way to dump it somewhere. I'll try to dump it into the file, though, since #include <cstdio> seems to compile ok...12:53
<bonniot>is it Sun's JVM that is used in FreeBSD? I'm surprised that they make their whole source code available (although surely under a restrictive license)12:57
<ArtemGr>Sun source is available to anyone registered in SDN (JDC). FreeBSD can't redistribute sources, they ask the user (admin) to download the sources from Sun.12:58
Although available is not the latest version12:59
<bonniot>i see
<ArtemGr>FreeBSD now have a license to redistribute binaries, though.
* arjanb joins13:02
<bonniot>hello arjan13:04
daniel do you have more time again?13:30
<bonniot>not 100% yet, but I can take some time once in a while13:43
<arjanb>bug #1090909 could be closed if the website is updated from cvs13:50
<bonniot>ok, i'll do a bit more cleanup and update it14:00
<CIA-1>03bonniot * 10Nice/web/manual.xml: Cleanups.14:01
<arjanb>some time ago i looked at disallowing covariantspecialization of foo(?A) foo(?B) where A<:B14:07
but the code in nice.tools.typing involves entering constraints and such which i'm not familar with
<bonniot>couldn't it be done syntactically? that is, don't use constraints. just check if both the parent and the child have '?' in their type14:10
i think there are already helper methods to test if a type is a "maybe" type14:11
<arjanb>well the check of A being equivalent to B is needed too14:14
<bonniot>isn't that already done?14:15
<arjanb>i think not14:17
i'll be back later...17:18
* bonniot leaves
<ArtemGr>Wow, JDK finally compiled and I got that string which is too long...17:42
<arjanb>how does that string start?17:44
<ArtemGr>I'll submit a bug report about InternalError and attach the string I obtained?
Here are first lines from the "string":
*S Default
1 userOperator.nice
2 symbol.nice
3 userOperator.nice
4 symbol.nice
5 userOperator.nice17:45
6 symbol.nice
7 userOperator.nice
8 symbol.nice
9 userOperator.nice
10 symbol.nice
11 typedef.nice
12 typedef.nice
13 typedef.nice
<arjanb>that looks like a byteattribute
<ArtemGr>What is byteattribute?17:46
<arjanb>i mean bytecode attribute
<ArtemGr>It looks to me like some information about the "bossa.syntax.fun" class, which probably have too many methods or too many of something else.17:48
<arjanb>this one is named SourceDebugExtension
<ArtemGr>After the numbered lines there is a lot of numeric information, like that:
2373 methodContainer.nice
2374 symbol.nice
2375 ai.nice
2376 symbol.nice
2377 ai.nice
<arjanb>it's line number information for debuggers
<ArtemGr>Oh, I see.
Should I submit this somewhere?
It's a 81596 bytes file.17:50
<arjanb>i think you should submit a bugreport to sun complaining about arbitary limit on the SourceDebugExtension attribute17:52
maybe we can add an option to nicec not to emit debug info as workaround17:53
can't the bytecode compactor you're using remove attributes?18:01
hmm it would be bugreport for Nice too because a workaround is needed18:13
no need to attach the actual string though18:14
the bytecode specification seems not to mention a limit on the attribute size18:16
*away for dinner18:17
* zzorn leaves18:24
* zzorn joins18:30
<ArtemGr>I'll try if obfuscator helps later, tomorrow maybe.18:39
I've submitted all information into the bug report.18:46
* ArtemGr leaves18:48
* bonniot joins19:51
arjanb: is 1113263 fixed?19:54
<arjanb>the verify error yes19:55
<bonniot>so cannot it be closed?19:56
<arjanb>ArtemGr has a point that having 'assert false' never returns is not consistent with the java semantics19:57
<bonniot>what is the point?19:58
i don't see this in the bug report (i didn't see the big testcase)19:59
oh and the -Xdebug bug in the jvm is caused by the SourceDebugExtension of bossa.syntax.fun being too large20:08
<bonniot>does it exceed a limit in the JVM specification?20:10
<arjanb>the spec doens't state any limit on attributes
haven't looked at jsr 45 though20:11
<bonniot>we should probably have a command-line option to disable source line information20:12
we could also see if the info could be put in a more compact way
the filenames part contains duplicates
<bonniot>i'm looking at 1090679 atn20:21
<arjanb>i tried adding maybe primitive variants of == and != but that caused specialication warnings20:23
<bonniot>i'll do it at the bytecode generation level20:26
<CIA-1>03bonniot * 10Nice/stdlib/nice/lang/inline/ReferenceOp.java: Optimization and cleanup.20:30
<arjanb>cvs is slow today, running > 15 minutes now :(20:45

Generated by Sualtam