This is the mail archive of the cygwin mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Java alternatives (was Re: Updated: stable compiler package gcc4-4.3.4-3)


Yaakov (Cygwin/X) wrote:

  Happy New Year, since then, BTW :) pardon me for taking a little time to reply.

> On 11/12/2009 14:58, Dave Korn wrote:
>> As part of its infrastructure, GCJ uses the Eclipse Compiler for Java
>> (ECJ)

> I still don't understand what the problem/worry is with licensing.  ECJ
> is under the EFL-1.0, an OSI-approved license.

  I just forgot to snip that from the text.  The only worry is self-hosting.

> As for self-hosting, I am finishing up a build of JamVM with GNU
> Classpath that can build and run ecj, ant and its dependencies,
> jdk6-langtools (Sun javac/javah/javadoc/javap), and even GUI toolkits
> SWT, Java-GNOME, and QtJambi.

  Brilliant, that will solve the only worry!

  BTW I'm going to ship AWT with the next GCJ, if that's of any interest to
anyone.  It builds and appears to be passing the mauve testsuite, so that's
probably worth chucking over the wall and letting people start playing with.

> To that end, I would like to work out with you a change to the
> alternatives(8) management in gcc4-java.  Since it is difficult to
> adjust launcher scripts for Java programs to figure out which Java
> interpreter/compiler to use at runtime, I suggest that programs supply
> symlinks to /usr/bin/java, javac, etc., as so:
> 
> Master: gcj-4 -> gcj
> Slaves: i686-pc-cygwin-gcj-4 -> i686-pc-cygwin-gcj, gcjh-4 -> gcjh,
> gij-4 -> gij
> 
> Master: gij-4 -> java
> Slaves: /usr/share/java/libgcj-tools-x.y.z.jar -> /usr/lib/tools.jar
> 
> Master: gjavah-4 -> javah
> 
> Master: gjar-4 -> jar
> 
> Master: gappletviewer-4 -> appletviewer
> Slaves: (Classpath tools) gjarsigner-4, gkeytool-4, gnative2ascii-4, etc.
> 
> All of these would be with a rank of 40.  Thoughts?

  (I don't understand the notation you're using there.  All the arrows point
in the opposite directions to what I'd expect for a symlink and there aren't
as many full paths as I'd expect to see either).  I've got no objection to
splitting up the single big monolithic set of alternatives that I currently
group under master "gcj" into a bunch of subpackages if that's what you're
saying?  If you'd like to rewrite the current postinstall/preremove scripts to
do what you want, that might be the most explicit way to show me what you mean
and I'll happily PTC them (by which I mean sure, anything sane I'll throw in
and use it!)

>> There are problems with the Java compiler that are less well understood,
>> but also manifest themselves as testsuite failures in the thread- and
>> process-related tests.
> 
> I see there are some things that JamVM/Classpath are able to run that
> GIJ cannot; my classpath patches in Ports SVN may be helpful.

  Thanks, will take a look.

> On an unrelated note,
> /usr/sbin/fix-libtool-scripts-for-latest-gcc-runtimes.sh in 4.3.4-3 does
> not have its executable bit set, so the postinstall attempt to call it
> fails (Permission denied).

  Argh, and thank you for pointing that out.  It appears that I lost all the x
perms on all the scripts in my last build, probably caused by my using win32
editors (followed by d2u) when adjusting them.  I've spotted that and fixed it
manually a few times but I'd better script it to avoid future accidents; it'd
be an easy thing to add to the cygport.

    cheers,
      DaveK


--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]