FYI, I've reported the jMock issue [1] and they already checked the fix in.
[1] http://jira.codehaus.org/browse/JMOCK-126
2007/5/21, Alexey Varlamov <[EMAIL PROTECTED]>:
Hi Alex,
2007/5/21, Alex Grigorovich <[EMAIL PROTECTED]>:
> Hi!
>
> I thought you may be interested: here's a recollection of problems I
> have encountered while trying to use Harmony for a small project at
> work.
Thanks for the feedback, we're quite interested indeed.
> First, jmock-based unit tests fail when run under Harmony's JVM.
> Compiler doesn't seem to matter: both sun's and harmony's work. I have
> attached a sample program which reproduces the following error:
[skipped]
This uncovers a bug in JMock2, matching reflection method instances
via identity operator ('==' vs equals()), see
jmock2/src/org/jmock/internal/matcher/MethodMatcher.java:18
I will report this issue to JMock developers until you beat me to it ;)
> Second, our java.nio-based network code which works on Sun's JVM
> throws a SockectExcepion on innocent-looking code:
> java.net.SocketException
> at
org.apache.harmony.luni.platform.OSNetworkSystem.select(OSNetworkSystem.java:312)
> at
org.apache.harmony.nio.internal.SelectorImpl.selectInternal(SelectorImpl.java)
> at
org.apache.harmony.nio.internal.SelectorImpl.select(SelectorImpl.java:167)
> at
com.shamrock.smf.niosmpp.net.SelectorHandler.run(SelectorHandler.java:156)
>
> These exceptions don't seem to affect the application though. They
> seem to happen after setting OP_WRITE on each interesting selection
> key. Relevant portion of source code looks like this (unfortunately I
> cannot provide a short runnable example):
>
> while (true) {
> try {
> /* register new channels etc. */
> synchronized(this.writableKeys) {
> /* for each key in writableKeys */
> if (key.isValid()) {
> key.interestOps(key.interestOps() |
> SelectionKey.OP_WRITE);
> }
> }
> selector.select(selectTimeout);
> Set set = selector.selectedKeys();
> /* process selected keys */
> set.clear();
> }
> }
>
> However, the problem might as well be in buggy application code. Any
> ideas on how to debug this?
>
I did not look closer here, we have a few decent networking experts on
the list, just wait them to speak up. BTW, do you mind to file a bug
report to JIRA for this issue?
>
> Third, VM hard-codes default character set to iso-8859-1 instead of
> using process' locale settings. Every UTF-8 user then has to manually
> set file.encoding system property, or otherwise get garbage on console
> when output does not fit in ASCII.
Yep, there is already a bunch of entries in our bug tracker depending
on this same issue, it must be fixed soon.
> Lastly, is there a native javac adapter for ant? One can use
> build.compiler=extJavac or write a custom CompilerAdapter, but a
> drop-in replacement for Sun's javac would be nice. Google even finds
> an implementation of com.sun.tools.javac.Main in harmony's subversion
> repo, but it seems to have been removed later (for trademark
> reasons?).
Sure, there is jdktools subproject which currently provides javac,
javah, keytool & jarsigner, and there is universal native starter for
all of them. The removed sources were mere duplicates between
"classlib" and "jdktools" subprojects.
Seems you've tried M1 binary snapshot, you've probably chosen a JRE
download? I believe JDK snapshots includes javac executable among
other tools...
> Harmony & OS details:
>
> $ java -version
> Apache Harmony Launcher : (c) Copyright 1991, 2006 The Apache Software
> Foundation or its licensors, as applicable.
> java version "1.5.0"
> pre-alpha : not complete or compatible
> svn = r533500, (Apr 30 2007), Linux/ia32/gcc 4.1.0, release build
> http://incubator.apache.org/harmony
>
> OS: Debian linux 4.0, kernel: 2.6.18-4-686 #1 SMP, libc-2.3.6
>
> Thank you.
>
> --
> Alexey Grigorovich
>
>
Stay tuned, Harmony advances in a good pace. So hope to hear from you
again once in a while ;)
--
Alexey Varlamov
