The 2007 JavaOne call for papers closes this friday. According to conference organizer Anette Vernon, there are currently only a little over 300 submissions so far, but appearently most people do submit at the last minute judging from earlier conferences.

New this year is that technical contents doesn't have to be 100% about the Java platform, but can be about for instance scripting, Open Source and community development, consumer technologies, or next generation web. This is one of the world's biggest developer conferences with around 14000 participants last year, so why not hurry up and submit your paper?

Edit: Still kicking around in the "pending" queue more than a week later, but since the deadline has closed I doubt it will be accepted....

My second attempt to do build the whole JDK, and to make life a bit exiting I am going to use a newer OS and gcc version than is recommended. I wouldn't do this if I was building a production ready runtime, but I am just trying to fix and verify a few starter bugs. I simply followed the directions here. Pretty straightforward, all required C source files are available to install through YAST.

When doing make dev-sanity to test system setup first time, I got the following errors.

ERROR: You do not have a valid GCC29_COMPILER_PATH setting.
    Please check your access to /java/devtools/linux/gcc29/usr/
    and/or check your value of ALT_GCC29_COMPILER_PATH.
    This will affect you if you build the plugin target.

ERROR: You do not have access to valid Mozilla header files.
              Please check your access to /java/devtools/share/plugin/mozilla_headers_18/java/bool.h
              and/or check your value of ALT_JDK_DEVTOOLS_DIR, ALT_MOZILLA_HEADERS_PATH,

ERROR: Missing ./../src/share/lib/fonts/LucidaTypewriterRegular.ttf.
              Verify you have downloaded and overlayed on the source area all the binary files.

To fix the font error, download the JDK 7 binaries jar, unpack with java -jar JARNAME and copy the files -r into the source tree.

Mozilla error, download Mozilla headers jar, unpack anywhere, export ALT_MOZILLA_HEADERS_PATH=${WHERE_YOU_UNPACKED_IT}/share/plugin/.

The last error is that GNU GCC egcs 2.91.66 (that EXACT version) is needed to build Mozilla java plugin. The installation instructions state, if you don't want to build the plugin, this is not needed. But it doesn't really say HOW to skip building the plugin. Well, if you already have an earlier installation, you can try this, which I found at Cay Hortmanns blog:
  export ALT_GCC29_PLUGIN_LIB_PATH=/usr/local/java/jdk1.6.0/jre/plugin/i386/ns7-gcc29

Next error I get was something like this:
>ERROR: You do not have access to the previous java release jre bundles.
>ERROR: You do not have access to the previous java release sdk bundles.

You can fix this by downloading these bundles and messing about with the variables ALT_PREVIOUS_RELEASE_PATH, ALT_PREVIOUS_JDK_FILE and ALT_PREVIOUS_JRE_FILE. Last time I tried that I never got it to work. But there is an easier way, and that is just to skip building the images target. And the solution to this I got from Tim Bell
Export the environment variable SKIP_COMPARE_IMAGES

Now I passed the make dev-sanity test. Horray! Now I go on to try "make dev".

Next error:
make[3]: Entering directory `/home/larsw/java/jdk7src/trunk/motif/lib/Xm/util'
make[3]: *** No rule to make target `/usr/X11R6/lib/X11/config/Imake.tmpl', needed by `xmkmf'. Stop.
make[3]: Leaving directory `/home/larsw/java/jdk7src/trunk/motif/lib/Xm/util'

Installing motif-devel headers through YAST fixed this.

Next error: ../../../../src/solaris/hpi/native_threads/src/interrupt_md.c:115: error: static declaration of 'sigignore' follows non-static declaration /usr/include/signal.h:377: error: previous declaration of 'sigignore' was here

Ivan Tarasov had the solution for this.

Next error: ../../../../src/solaris/bin/java_md.c: In function 'get_cpuid': ../../../../src/solaris/bin/java_md.c:1119: error: PIC register '%ebx' clobbered in 'asm'

If my knowledge of C is weak, when it comes to assembler it is non-existant. :-)
However, both Cay and Ivan get by this by commenting out the clobbering of the ebx register. I join the line of those waiting to hear from the experts if this is an acceptable solution. The code comments just say: "ebx is callee-save, so push it even though it's in the clobbers section". Doesn't feel good to comment away stuff I don't understand until it works, but for now I guess I'll just put a book on assembler on my Christmas wish list. I hoped to get some sleep during the holidays, but there is always next year I suppose...

Next error:
t/JTop/JTop/classes ../../../../src/share/demo/management/JTop/JTop.java ../../../../src/share/demo/management/JTop/JTopPlugin.java ../../../../src/share/demo/management/JTop/JTopPlugin.java:53: package com.sun.tools.jconsole does not exist
import com.sun.tools.jconsole.JConsolePlugin;

First attempt: In the file j2se/make/mkdemo/Makefile I removed management and scripting as targets. They are only demos after all.

Second attempt: After looking around, I found this! Do NOT export J2SE_ONLY, it is an old, probably non-working, option! I tried to build as small as possible of course to make it easy, but it was that which caused management/JMX to not build, but these are needed later. Hopefully this option will be removed.

Final errors: /bin/cp: cannot stat `/usr/local/java/jdk1.6.0/jre/plugin/i386/ns7-gcc29/ns7/libjavaplugin_oji.so': No such file or directory /bin/cp: cannot stat `/usr/local/java/jdk1.6.0/jre/plugin/i386/ns7-gcc29/libjavaplugin_nscp_gcc29.so': No such file or directory

These are caused by bugs(?) in the Makefiles. When I Googled for it, I found that once again Ivan had the solution.

Now it all builds, for real this time.

This will also be the last entry in this series, since I am going to remove Suse from all my home and work computers until Novell gets a clue. I guess my next distro will be Ubuntu. I really really like KDE, but Kubuntu doesn't feel as polished yet. Hmm...maybe Fedora?

New article submission by me:

It is now official, Sun is releasing all versions of Java (Standard, Enterprise, Micro) under GPL version 2.0. With around 6 million lines of code in Java SE alone, this is probably one of the biggest transitions to Open Souce ever. There will be webcast from Jonathan Schwartz and Rich Green at 9:30PT.

Edit: Someone beat me to it.

Submitted story:

The current issue of The Economist has its science supplement Technology Quarterly available online for free. In this issue, articles discuss among other things vat grown clone meat, the (impossible) economics of fuel from trash, episodic gaming, a Chinese eco-city, high tech concrete and computer "mind reading" through Bayesian engine augmented cognition .

End submission.

Slashdot journal is buggy by the way. In my first post I have tried to enable comments several times, either through editing the entry and selecting "comments enabled" in the dropdown menu, or clicking on the "enable comments" link. Neither works.

Edit: Rejected.

How that JDK is being open sourced, I thought it would be interesting to build JDK7, so here are some notes from my experiment in case anyone is interested. I shall try to submit bugs that I find whether they are in Linux, Netbeans or Java, and if my programming skills are up to it I will try to solve them.

Day 1: Latest Netbeans (Relase Candidate 1) downloaded. Needs Java 5 to start, but I have (yet again) problems with Yast in Suse 10.1, it instantly crashes whenever I try to update anything, so I'll just download JDK6 from Sun manually.

Once JAVA_HOME is set, Netbeans starts. Notably faster startup on JDK6, Swing is snappier now, plus very nice anti-aliased fonts. Swing is defenitly catching up with SWT quickly. Subversion plugin downloaded and installed from within Netbeans without any problems. No "checkout from svn/cvs as new project" that I could see, this is something that I'm missing from Eclipse. The subversion menu has a "checkout" command though, it is the only enabled menu command, so I'll guess I try that, and try to import the sources as a new project once it is done.

Repository location URLs for JDKs are a bit hard to find, but I finally found it at this blog. You will need a username and password to do a checkout for now, but this will hopefully change as the open sourcing process continues. The Netbeans checkout wizard fast and easy, but once checkout starts, there is only no svn output feedback by default, just a "process running" animation. I have to go the destination dir to see the files appear, not very good, if download grinds to a halt, will I be able to see it? I have a very slow DSL (damn you, Spray!), so I leave the computer alone for an hour. When I come back, Netbeans is appearently still running the checkout process, but repeated "df" command from bash shows that the download must have halted, nothing is appearing on the drive. Just what I worried would happen. Netbeans should either add svn output so we can see what is happening, or a timeout.

I kill the Netbeans checkout process, and try to find a way in the preferences to get a visible svn output in Netbeans, but no luck. I go to command line instead, clean up the old locks, and try running an update. Seems to work, and now I see progress for each file. Seems command line is the way to go... for now at least, we shall see if it works better once a project is imported in Netbeans. End day one.

From now on I'm going to journal all my submitted stories, so that next time the Slashdot editors say they only post stories by Ronald Piquepaille, beatles-beatles et al is because they don't get any good submission by us others, I'll have something to point to and ask the Slashdot public what they think. :-)

Submitted today:
As Sun is working towards open sourcing the JDK, they have previously asked the community which bug tracker to use, and now Mark Reinholt (Chief Engineer for the java platform) blogs about which distributed source-code management tool to use. They are currently leaning towards the Python based Mercurial, but Linus Torvald's Git and Bazaar are other solutions being evaluated.

Edit: Rejected, half a day later.