Re: Disabling Kaffee
Stefan Bodewig apache.org> writes: > > Hi, > > the Kaffe build doesn't get very far and has ill side effects on the > test Gump run (occasionally it terminates a test run running at the > same time). Makes sense to me, too, as gump wasn't tracking Kaffe's cvs head, and it has changed quite radically internally since we set up the gump instance. I've got a 'setting gump' up for kaffe/your own vm thing written up somewhere, I'll try to get it out on my blog someday. It's not very hard. Thanks for everything the gump devs have done to make java libre happen, you rock. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [Kaffe] Has the rmic entry point moved?
Stefan Bodewig apache.org> writes: > > On Wed, 9 Mar 2005, Dalibor Topic kaffe.org> wrote: > > Stefan Bodewig apache.org> writes: > > >> The build failure indicates that Ant doesn't find either > >> gnu.java.rmi.rmic.RMIC or kaffe.rmi.rmic.RMIC. > >> > >> Has the class moved/been renamed or have I broken Ant lately? > > > > Thanks for reminding me, Stefan! I've moved them yesterday. > > OK, Ant should now know about it. kaffeh is still there, I hope. > > Maybe Steve should have addressed you with his open letter > <http://www.1060.org/blogxter/entry?publicid=7E29E4A93CCEDA5E46DA7A87D8167168> > as well Ah, yes. Guilty on all accounts. :( I guess the design decision to use 'unofficial' entry points is now set in stone in ant? I kind of hoped ant would eventually just invoke the scripts/native code/etc. directly without going through the JVM first. But Steves blog means that it is pretty unlikely to happen, as that would certainly force a lot of people to double-check their builds. :( I'll probably add a tiny tools.jar to shut up ant's 'no tools jar' warning and make it run javac and rmic fine by default with Kaffe through the com.sun. entry points from Thomas' java-gcj-compat wrapper. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [Kaffe] Has the rmic entry point moved?
Stefan Bodewig apache.org> writes: > > Hi, > > we finally had a Kaffe run that got to the point where it tries to > build log4j or lucene again. The build failure indicates that Ant > doesn't find either gnu.java.rmi.rmic.RMIC or kaffe.rmi.rmic.RMIC. > > Has the class moved/been renamed or have I broken Ant lately? > > Stefan > Thanks for reminding me, Stefan! I've moved them yesterday. I've merged some more code from GNU Classpath tools project, and replaced our own tools, where possible to use the commonly maintained ones. In particular, the entry points for * serialver (gnu.classpath.tools.serialver.SerialVer) * rmic (gnu.classpath.tools.rmi.rmic.RMIC) * native2ascii (gnu.classpath.tools.native2ascii.Native2Ascii) * javap (gnu.classpath.javap.JavapMain) have been changed. The code is GPLd in cp-tools which means it needs to be invoked as external process, I guess. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Gump ApacheCon Europe?
Stefano Mazzocchi apache.org> writes: > > Dalibor Topic wrote: > > Yeah, I was thinking about submitting a 'how I learned to stop >> worrying and love > > the gump' > > great title heh :) > > great talk by leo at fosdem btw. i believe green banned it all on >> video, and > > puts it online somewhere. > > URL? > None so far. According to green's blog from http://www.spindazzle.org/green/index.php?p=39 it's not going to be oscar-worthy, though, thanks to RyanAir losing his luggage. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Gump ApacheCon Europe?
Adam R. B. Jack apache.org> writes: > > Questions, comments, suggestions? > > Gump hacking (by most) at such a macro and community level. I'd love to see > content simply on the people and endeavours that have been brought together > by it's processing on various platforms (JDK1.5, Kaffe, etc.) Yeah, I was thinking about submitting a 'how I learned to stop worrying and love the gump' thingie describing the experiences I had with the gump/kaffe collaboration and spreading praise on the gump team and the ASF in general :) great talk by leo at fosdem btw. i believe green banned it all on video, and puts it online somewhere. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [Kaffe] bootstrap-ant seems to think Kaffe is JDK 1.5
Stefan Bodewig apache.org> writes: > > ... since optional/condition/IsPingable.java shouldn't be compiled > otherwise. > > Ant checks for the class java.lang.Readable to detect JDK 1.5, has > this recently been added to Kaffe? Yeah, just yesterday. Merge of a few non-generics bits from the generics branch within GNU classpath. I guess the java.lang.management classes are a safe bet, though :) cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: sync has problems with symlinks
Stefan Bodewig apache.org> writes: > > On Wed, 09 Feb 2005, Stefan Bodewig apache.org> wrote: > > > (2) This symlink should get removed in the tearDown method of the > > corresponding tests and that seems to fail on Kaffe. To validate > > this, I'll remove the ant working copy again and watch the test-ant > > build on the next Kaffe run closely. I'll also add some diagnostic > > output to Ant's test itself. > > Kaffe didn't manage to bootstrap Ant for some unknown reason (the > class it fails to find is there). If anybody could grab the log of > "test-ant" when Kaffe manages to get there the next time, that would > be nice. Hmm, this appears to be fixed now, though I don't recall checking in a specific fix for it into Kaffe. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: sync has problems with symlinks
Stefan Bodewig apache.org> writes: > > Hi, > > the second time in a week I had to clean out the Ant working copy in > the Kaffe workspaces because it had a stray symlink in it. > > Somewhere during the tests, Ant creates symlinks inside the source > tree and the sync step fails to remove them - which then in turn makes > Kaffe try to compile classes that should be excluded. > > I'll try to find a solution for the Ant side, but if somebody with > more experience in Python could look into it, it would be great. > > sync thinks the link was a directory and the removetree call fails. > My guess is that we need to do lstat instead of stat and try a plain > remove on links (just like files), but I may be completely wrong. I have seen it but have't looked into it yet. I'm not sure from your description whether you think it's a bug in Kaffe, Gump, Python or Ant :) Could you elaborate? We talked about it a bit on IRC. Adam Heath fixed a symlink problem in Kaffe's libraries a while ago, chances are it resurfaced. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [Kaffe] current issues
Stefan Bodewig apache.org> writes: > > On Fri, 28 Jan 2005, Stefan Bodewig apache.org> wrote: > > > I've found kaffe.tools.native2ascii.Native2ASCII in Kaffe's CVS and > > plan to adapt the Ant task to it. > > This should be done now. Dalibor, I couldn't find any class that > looked like an imported Classpath version of it, just ping me when you > are done, please. Yeah, sorry about the confusion. It turned out that cp-tools doesn't have Native2ASCII yet, so Ito will submit our implementation to GNU Classpath. Chances are it wil be renamed by cbj to something to match GNU Classpath tools projects naming scheme. When that happens, I'll make sure to ping you about it. > > Later in the build, we'll need as well. > > Making a facade task and supporting kaffeh is next on my list > of Ant/Kaffe Interop issues. Thanks! cheers, dalibor topic, still having to finish off rmic tonight. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [Kaffe] current issues
Stefan Bodewig apache.org> writes: > > Hi all, > > our major problem seems to be rmic ATM. I've thrown in an > unconditional -verbose when Ant invokes rmic, maybe the verbose output > of Kaffe's rmic is going to get us further. I'm hacking on fixing Classpath's RMIC. I've got it mostly working in my local tree, but had to do some hacking on the Kaffe build system last week in between :( It should be finished tonight. > The new java-service-wrapper project I've added shows some interesting > additional tasks that won't work on Kaffe. Right now it fails because > is only supported on Sun VMs, but I've found > kaffe.tools.native2ascii.Native2ASCII in Kaffe's CVS and plan to adapt > the Ant task to it. That would be quite cool, but I think I'll simply merge in the native2ascii from cp-tools (GNU Classpath tools) project tonight, so that you have one less legacy class name to maintain ;) > Later in the build, we'll need as well. So far I haven't > found the Kaffe version of it, is there one? Currently, it's the native code kaffeh executable. I plan to switch to cp-tools javah in java as soon as the merge with Classpath is finished (a few weeks). cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: derby build
Stefan Bodewig apache.org> writes: > By now you can rely on Gump running on JDK 1.4+. I'm not sure which > version of JDBC Kaffe ships with, though. According to http://www.kaffe.org/~stuart/japi/htmlout/h-jdk14-classpath.html we have all the methods & classes from java.sql and javax.sql form 1.4 implemented. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [kaffe] JUnit passed, but javadoc generation failed
Stefan Bodewig apache.org> writes: > > Hi, > > on the last run we've successfully built JUnit, which probably means > more of Swing is available to Kaffe now. > > Looking into junit's build log, I see that javadoc failed because > Kaffe's version of javadoc doesn't seem to be compatible with Sun's > command line options. > > Dalibor, can you provide me with a list of supported javadoc options? > Also, currently Ant's build file looks for the class > kaffe.util.NotImplemented - is this reliable or should we look into > java.vm.vendor instead? Thanks a lot for the bug report, it should be fixed now. I've taken a pretty long time to resync with the respective updates from gjdoc CVS head because it needs antlr now, and I had to figure out how to work it into our build system first. So, you'll need tp apt-get install antlr on brutus to make sure antlr.jar is there, and to enable gjdoc support explicitely in kaffe's configure line using --enable-gjdoc cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [Kaffe] Lucene fails in
Stefan Bodewig apache.org> writes: > > On Tue, 14 Dec 2004, Dalibor Topic kaffe.org> wrote: > > Davanum Srinivas gmail.com> writes: > > > >> > >> i think so... > >> > > > > Yeah, the RMIC code essentially the same as in old Kaffe, but under > > the GNU namespace. > > Unfortunately we don't seem to get to rmic right now. The Process bug bit Kaffe again. I'll look at the classpath class again. > Which version of Kaffe has the new package name? This is for the > ChangeLog I have to write, you know. sine 1.1.2, according to http://www.kaffe.org/cgi-bin/viewcvs.cgi/kaffe/libraries/ javalib/gnu/java/rmi/rmic/RMIC.java cheers,a dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [Kaffe] Lucene fails in
Davanum Srinivas gmail.com> writes: > > i think so... > Yeah, the RMIC code essentially the same as in old Kaffe, but under the GNU namespace. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Problem running Apache Gump [brutus-kaffe]
Adam R. B. Jack apache.org> writes: > It isn't something we made configurable, although we ought have. I'll add > that. Thanks, Adam! cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Problem running Apache Gump [brutus-kaffe]
Adam R. B. Jack apache.org> writes: > > > > > Hmmm, what's wrong with the kaffe build? it keeps on sending these > messages. > > > It keeps getting this message: > > [gump brutus] /usr/local/gump/kaffe/gump $ svn update > svn: Failed to add file 'gump': object of the same name already exists > > Ever since I had to delete/re-create the LIVE branch we've seen this. I do > an "svn cleanup" and then it works, but it seems it comes back. Perhaps it > is time to move kaffe/gump to kaffe/gump-svn-corrupt, re-checkout and > transfer any files over. What I don't get though is that a kaffe build > appears to be ongoing. Maybe it is some dross that needs to be cleaned > out/killed (and maybe it is part cause of the SVN problem.) speaking of killing runaway kaffe processes, would it be possible to get the default kill time for runaway processes down from 1h to 30 minutes? brutus seems to be a sporty enough machine for that (at least from kaffe's side) to make the single builds happen quickly enough (i hope, at least), and the hangs from kaffe (yep, our bug, i'm working on it) slow the build log down quite a bit. I can assume that the hangs will increase as more projects become buildable, so while we're busy finxing them, please consider setting the kill time lower for kaffe. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [xalan][kaffe] Xalan build with Jikes uses classes in rt.jar
Stefan Bodewig apache.org> writes: > > On Tue, 14 Dec 2004, Davanum Srinivas gmail.com> wrote: > > > not yet...right now am trying the new jikes compiler. > > I'm rather pessimistic. > > It is not really jikes' fault. It gets to compile If we are talking about jikes erranously getting confused about deprecation warnigs and treating them ocassionally as errors, I believe that was a bug in jikes and is supposedly fixed in jikes CVS. See http://www-124.ibm.com/developerworks/bugs/ ?func=detailbug&bug_id=4059&group_id=10 (sorry for split url) cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [Kaffe] Lucene fails in
Stefan Bodewig apache.org> writes: > > This is a question for Dalibor, I guess. > > It seems as if Ant didn't find the Kaffe rmic version it claims to > support. Ant should load the class kaffe.rmi.rmic.RMIC. After the merge of RMI support with GNU Classpath, that class has moved into the gnu.* namespace. It is gnu.java.rmi.rmic.RMIC now. > Is this the correct name? Which jar is supposed to contain it? It is in Kaffe's rt.jar on its bootclasspath. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [kaffe][status] jakarta-regexp & junit
Davanum Srinivas gmail.com> writes: > > Latest Biggies are jakarta-regexp and junit (see > http://brutus.apache.org/gump/kaffe/project_todos.html) > > jakarta-regexp: > build.xml copies source files from jakarta-regexp/src to > jakarta-regexp/build/src and it applies filters during the copy > operation. Kaffe fails to do this properly. Result being the copied > files are empty. If i remove the filter flag in build.xml, it works > fine. So this is on dalibor's and kaffe team's plate and it's easy to > replicate (just get jakarta-regexp cvs and run ant) Thanks a lot for helping so greatly with the bug chase! This particular bug should be fixed now in CVS head by Adam Heath. > > junit: > there are 2 problems here. the first one is the same as above, > copy/filter fails...in this case junit.runner.Version file is empty. > The other problem is an ongoing problem of CLASSPATH not completely > supporting all api's in javax.swing. Again, both problems are on > dalibor's and kaffe team's plate. Thomas Fitzsimmons checked in a preliminary version of it on the gcj gui branch, and I'll copy it over to kaffe tomorrow. Michael Koch might have the missing ingredients for JTree finished, too. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [kaffe][status] jakarta-regexp & junit
Davanum Srinivas gmail.com> writes: > > Stefan, > > i can see sun.boot.class.path in http://brutus.apache.org/gump/kaffe/ > Yep, that's the one. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Kaffe/Jikes builds
Stefan Bodewig apache.org> writes: Hallo Stefan, > On Wed, 8 Dec 2004, Dalibor Topic kaffe.org> wrote: > > > the Xalan failure looks a bit bizarre, as I'd expect it to actually > > use the org.w3c classes from xml-apis.jar. > > Which it seems to do in the JDK 1.5 builds (otherwise it would have > the same problems there as well). > > We use -Xbootclasspath for xml-apis to ensure our version gets used, > maybe this doesn't work for Kaffe? Or just not using Jikes? I'll > probably need some -verbose output to be sure. I'll give it a look later, too. > > Beside the lacking methods in class library, the major thing that > > kills builds seems to be an oddity in jikes. See > > http://brutus.apache.org/gump/kaffe/jakarta-commons/commons-lang/gump_work/ > > build_jakarta-commons_commons-lang.html for an example: > > Why don't I see any result pages on brutus ATM? Ahh, the last Ant > build took more than an hour and was killed. Any idea why? It seems to be some issue with hanging processes, I'll try to merge in the process implementation from GNU Classpath into Kaffe to see how well that works. > Maybe turning off -deprecated will help. I'll give it a try. Thanks. > > This > > http://brutus.apache.org/gump/kaffe/cpptasks/ant-contrib-cpptasks/gump_work/ > > build_cpptasks_ant-contrib-cpptasks.html > > Page is not there, but any Ant build will certainly need some kind of > XML parser. I'll add Xerces. Sorry about that, I had to split the URLs because the gmane web posting HTML widget hates lines longer than 80 chars ... cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Kaffe/Jikes builds
Stefan Bodewig apache.org> writes:
>
> Hi,
>
> I include Jikes in the subject since the Ant breakage is due to a
> Jikes/Javac incompatibility (jikes won't accept -source 1.2).
>
> The Ant build should work on the next Kaffe run.
Thanks a lot Stefan! It worked pretty well, and more projects have now built.
> For the rest of the problems it seems that Kaffe's classlib doesn't
> implement enough of Swing (without JUnit we won't get very far) and at
> the same time implements too much of DOM3 (Node#getUserData not
> implemented in Xalan, yet).
the Xalan failure looks a bit bizarre, as I'd expect it to actually use the
org.w3c classes from xml-apis.jar.
The warning from jikes
[javac] *** Semantic Warning: I/O warning: "No such file or directory"
while
trying to open
/home/gump/workspaces2/kaffe/workspace/xml-xalan/java/${build.boot.class.path}.
seems to indicate that kaffe somehow fails to turn ${build.boot.class.path} into
something useable.
I don't know what build.boot.class.path is though, nor was google helpful in
really finding out ;)
Beside the lacking methods in class library, the major thing that kills builds
seems to be an oddity in jikes. See
http://brutus.apache.org/gump/kaffe/jakarta-commons/commons-lang/gump_work/
build_jakarta-commons_commons-lang.html
for an example:
[javac] Found 2 semantic errors compiling
"/home/gump/workspaces2/kaffe/workspace/jakarta-commons/lang/src/java/org/
apache/commons/lang/builder/StandardToStringStyle.java":
[javac] 81. public boolean isShortClassName() {
[javac]^^
[javac] *** Semantic Error: The overridden method "boolean
isShortClassName();" is deprecated in type
"org.apache.commons.lang.builder.ToStringStyle".
[javac]102. public void setShortClassName(boolean shortClassName) {
[javac] ^---^
[javac] *** Semantic Error: The overridden method "void
setShortClassName(boolean shortClassName);" is deprecated in type
"org.apache.commons.lang.builder.ToStringStyle".
I guess that's a bug in jikes.
This
http://brutus.apache.org/gump/kaffe/cpptasks/ant-contrib-cpptasks/gump_work/
build_cpptasks_ant-contrib-cpptasks.html
looks like a missing Xerces on the classpath for jikes.
cheers,
dalibor topic,
off to bed for today
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
Re: Kaffe build
Davanum Srinivas wrote: dalibor, What should we try next? any ideas? we've been stuck here for a few weeks now (http://brutus.apache.org/gump/kaffe/ant/bootstrap-ant/gump_work/buildscript_ant_bootstrap-ant.html) thanks, dims Thanks for the reminder, Davanum! I can't seem to be able to reproduce it with current Kaffe from CVS head, I'll give it a go with kaffe in unstable at home a bit later. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: kaffee local-env-brutus.sh on brutus
Adam R. B. Jack wrote: That looks like the same problem I just had, although I was running the gump.py in the root directory rather than the cron directory. I believe you need to run gump.sh from within the cron directory (cd cron && ./gump.sh) since it uses the gump.py in the current directory. Yeah, sorry folks -- it's a bit of a mess. We've had all sorts of good intentions, and folks starting command line version of things (verse cron versions), and we need to clean up and focus until we have time to do them. Running the gump.sh in ./cron ought do the trick. Thanks a lot for the quick reply, that does the trick indeed. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: kaffee local-env-brutus.sh on brutus
Stefan Bodewig apache.org> writes: > > On Fri, 12 Nov 2004, Davanum Srinivas gmail.com> wrote: > > > Now it's showing this: > > The jikes warnings can be ignored (though jikes is right and Ant's > source should be fixed). > > Right at the end it looks as if Kaffe dies while running the freshly > boostrapped Ant. Looking at > <http://brutus.apache.org/gump/kaffe/ant/bootstrap-ant/index.html> it > says "Elapsed: 60mins" - so probably Gump has killed it. > > Stefan > I've tried to look into it, but my own gump instance seems to be broken after an update for some reason :( I'm getting: $ bash cron/gump.sh build _ | __|_ Apache_ ___ | | | | | | . | |_|___|_|_|_| _| |_| ~ v. 2.2 ~ Traceback (most recent call last): File "gump.py", line 118, in ? main(sys.argv) File "gump.py", line 108, in main getattr(command, "process")(options,arguments) File "/var/tmp/topic/gump/python/gump/commands/build.py", line 41, in process workspace = WorkspaceLoader(options.isCache()).load(ws) any idea what's going on? cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: The Kaffe instance and ant-bootstrap
Stefan Bodewig apache.org> writes: > > On Tue, 2 Nov 2004, Niclas Hedhman hedhman.org> wrote: > > On Monday 01 November 2004 21:12, Dalibor Topic wrote: > > > >> Since you're using the normal debian packages, setting a few > >> additional env vars should do the trick for the bootstrap with > >> kaffe: > >> > >> extras for ant bootstrap with kaffe > >> ## > >> ## Use jikes with kaffe's class libraries on bootclasspath > >> export JAVAC=jikes-kaffe > >> > >> ## Jikes doesn't like javac.source=1.2 in build.xml, but only > >> ## accepts 1.3 or 1.4 export ANT_OPTS="-Dbuild.compiler=jikes > >> ## -Djavac.source=1.3" > > > > So, are you suggesting that we do the above, or are you suggesting > > that we should prepare to get Kaffe from CVS? > > Only that we (well, anybody with root privileges) install the > appropriate jikes package and set the two env variables before > starting the bootstrap script. Yep. Getting & building Kaffe from CVS is probably too much manual labour if you don't intend to hack on Kaffe specifically, while anyone can update the installed versions of kaffe and jikes-kaffe using apt, I guess. And I can take care of injecting bug fixes from/for gump into Kaffe's package on debian, being on the pkg-kaffe team there ;) > The workspace will need to set the build.compiler property to jikes > for all Ant builds as well. Yep. Is setting the environment variable ANT_OPTS enough for gump's ant builds, or does one need to put it somewhere special, too? cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: The Kaffe instance and ant-bootstrap
Stefan Bodewig apache.org> writes: > > On Sun, 31 Oct 2004, Dalibor Topic kaffe.org> wrote: > > Stefan Bodewig apache.org> writes: > >> > >> Ant's bootstrap script can use jikes easily by setting the JAVAC > >> environment variable. > > > > Ah, cool, I didn't know that. Since which version of ant is that the > > case? > > Hmm, the change has been made four years ago for the jpackage folks > (who wanted to bootstrap using jikes . So it in there since > Ant 1.3. > > <http://cvs.apache.org/viewcvs.cgi/ant/bootstrap.sh?r1=1.22&r2=1.23> Thanks, I should have just read the source :) > > Yes. the main reason is that kjc is not greedy wrt to resolving > > missing classes, whereas most other compilers are, so people > > naturally write build.xml files that don't work with kjc, which > > needs to be fed all classes on the command line. > > Detecting those builds may be interesting in itself, but folks would > be unlikely to change their build files - but rather tell you that > kopi was broken. Yeah. I guess both the greedy and the minimalist approach have their pros and cons, but the greedy approach has become so common, that it wouldn't be likely that someone would maintain the build for the minimalist approach, I assume, unless it offered some major breakthrough in compiling Java code :) Otoh, it may be interesting to write a 'greedyness' wrapper for compilers in ant, rather then implementing greedy dependency handling in each minimalist compiler[1] separately. If Ant maintains dependency information, it could use that to generate transitive dependency closures of sources needed to build, right? cheers, dalibor topic [1] kjc, gcj, possibly ikvmc[2], jc's genobj[3], and other java-to-something-unusual compiler tools. [2] http://www.ikvm.net/userguide/ikvmc.html [3] http://jcvm.sourceforge.net/doc/jc.html#genobj - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [proposal] removing non-ASF leaves from the workspace
Niclas Hedhman hedhman.org> writes: > > On Monday 01 November 2004 17:00, Leo Simons wrote: > > Kaffe is very much a leaf not a dependency (I know no ASF project > > that can only be built using Kaffe), yet using it for experimental > > runs doubles the amount of cpu and disk space used. > > For the record, there are 8 attempts at starting a Gump build every day. > 1 is Kaffe, 1 is JDK1.5, 1 is 'test' and the others are the official build. > So, it is not completely accurate to say that the Kaffe build instance doubles > CPU/disk resources. My long term (a few weeks, I hope) plan is to have a second gump instance on developer.classpath.org with the free runtimes to both take the extra load off ASF, and to give further free runtimes like gcj, IKVM, JamVM, etc. a go at building the free java software world from scratch. > > While I appreciate the goal of being able to have a truly free java > > stack and how using Kaffe to build ASF projects helps towards attaining > > that goal, we're also doing "public service" towards the GNU people in > > this way. > > Looking at the fact that a Kaffe developer (dalibor) has taken interest in > Gump, installed his own instance and trying hard to get things going, is IMHO > a good testament to the appreciation of Gump. I appreciate in particular the extremely helpful developers on #gump. Thanks a lot for those that helped me set myself up. I still have to pay you back with the wiki page, though ;) Seriously though, I think gump is just wonderful, and is going to help leapfrog the whole free runtimes thing quite a bit as a side effect. Today, there is a lot of great free software written in Java, not least thanks to ASF's successful projects. Unfortunately, there are often some small issues that prevent some free Java softare from running on free runtimes. Gump can help find and fix those small issues, as it did for GNU JAXP's small bug today for me. The small issues may be bugs in the free runtimes, or code that is just a little bit outside the letter of the specification, but runs fine on non-free runtimes. In particular the latter case is hard to anticipate a priori in a class library implementation. That's where having a gump can be very, very helpful: it shows what idioms other developers expect to work. Finally, I see great potential in maintaining a confidence level in a free java stack, once we're there ;) That could help quite a bit wrt to packaging efforts of free software written in Java for operating systems like GNU/Linux, *BSD, etc. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: The Kaffe instance and ant-bootstrap
Dalibor Topic kaffe.org> writes: > > I've had to solve a similart problem for bootstrapping ant with kaffe & jikes in > kaffe-extras module of kaffe's CVS. There I've patched ant's bootstrap scripts > for that [1] but of course setting env vars is more elegant ;) And I can confirm that it works for me with Kaffe's CVS head. Since you're using the normal debian packages, setting a few additional env vars should do the trick for the bootstrap with kaffe: extras for ant bootstrap with kaffe ## ## Use jikes with kaffe's class libraries on bootclasspath export JAVAC=jikes-kaffe ## Jikes doesn't like javac.source=1.2 in build.xml, but only accepts 1.3 or 1.4 export ANT_OPTS="-Dbuild.compiler=jikes -Djavac.source=1.3" The next issue I had was a bug in GNU JAXP's lookup of SAX parsers. Thanks to gump for making it easy to spot! ;) With the issue fixed in my local kaffe tree, ant bootstraps fine and dies on ant build die to missing configuration bits: xml-apis, and xml-xalan. That's my next target ;) cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: The Kaffe instance and ant-bootstrap
Stefan Bodewig apache.org> writes: > > On Wed, 27 Oct 2004, Niclas Hedhman hedhman.org> wrote: > > > I have looked at the Kaffe instance and its inability to create the > > Ant bootstrap. > > > > The first level of problem is that the Kaffe compiler doesn't imply > > any source files that are not specified, and that results in the > > immense number of "Cannot find class". > > I think the only viable solution is to drop kopi and use jikes as a > compiler. The very same problem is likely to come up in javac tasks > later on as well. Yes. > There already is an issue in Jira for this. > > Ant's bootstrap script can use jikes easily by setting the JAVAC > environment variable. Ah, cool, I didn't know that. Since which version of ant is that the case? > > I don't know if we should proceed by convincing Ant to introduce > > this to their codebase, or we should maintain a separate bootstrap > > script. > > I'd prefer to really not use kopi at all. I've been told that even > the Kaffe people prefer to use Jikes. Yes. the main reason is that kjc is not greedy wrt to resolving missing classes, whereas most other compilers are, so people naturally write build.xml files that don't work with kjc, which needs to be fed all classes on the command line. > Any approach like yours will need to be maintained when classes get > added to Ant. Sorry for the late reply, I've been busy with some other issues, unfortunately. I'll look into making the bootstrap work with jikes & kaffe in debian by setting the JAVAC env var. I've had to solve a similart problem for bootstrapping ant with kaffe & jikes in kaffe-extras module of kaffe's CVS. There I've patched ant's bootstrap scripts for that [1] but of course setting env vars is more elegant ;) cheers, dalibor topic [1] http://www.kaffe.org/cgi-bin/viewcvs.cgi/kaffe-extras/patches/ ant-kaffe-1.1.3.diff?annotate=1.2 - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [RT] two wild thoughts
Stefano Mazzocchi apache.org> writes: > one thing I would love to do is to have gump bootstrap kaffe, jikes and > then bootstrap GNU classpath and then the apache stuff. > > it would be very sweet. > > would you guys be interested in helping out? Sure :) I've asked Jim Pick to post to this thread with his GUMP/Kaffe experiences. cheers, dalibor topic - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [RT] two wild thoughts
Stefan Bodewig apache.org> writes: > > Hi, > > Maybe you should have used two RT threads with meaningful subjects? > > On Mon, 09 Aug 2004, Stefano Mazzocchi apache.org> wrote: > > > Wild Thought #1 > > > > We now have Mono installed on Brutus. > > The Ant tests show that it doesn't work very well, in particular it > fails on stuff that works perfectly well in my installations. So > maybe Debian's Mono package really isn't ready for prime time, yet. > > > We just need to download the ikvm[1] and try to run our java tools > > on top of a "virtual" java virtual machine running on top of Mono. > > +1 once Mono really works on Brutus. We've had something like that for Kaffe[1] for a brief moment made by Jim Pick. See http://www.kaffe.org/~jim/kaffe-gump/ for the scripts, and http://www.kaffe.org/~jim/gump/log/ for the output. Jim has offered to host devel.classpath.org on a beefed up kaffe.org server, and we may have a gump setup there, too, eventually. cheers, dalibor topic [1] Always use the CVS head, we're too busy hacking to do releases ;) - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
