JDK specific classpaths

[Herve Quiroz]

Hi Greg,

On Tue, Apr 05, 2005 at 09:44:08AM -0600, Greg Lewis wrote:
> On Tue, Apr 05, 2005 at 03:32:05PM +0200, Herve Quiroz wrote:
> > On Tue, Apr 05, 2005 at 11:07:32PM +1000, Sam Lawrance wrote:
> > That said, and until some volunteer starts working on improving
> > javavmwrapper, I would go with method #2 (overridable JAVA_HOME).
> 
> The problem is that although javavmwrapper exports JAVA_HOME to its
> children, we're wanting to use JAVA_HOME during its invocation.
> 
> While I don't think a full blown tool to print out JAVA_HOME is
> necessarily warranted, it may be useful to have javavmwrapper do
> some substitutions on the command line arguments it is given, e.g.
> "s/%%JAVA_HOME%%/${JAVA_HOME}" in this particular case.  Presently
> I can only see uses for JAVA_HOME and the version, so this should
> be quite simple to implement.

I agree with you here.

I had already thought that a JAVA_HOME print out tool would have been
useful for Ant, which requires JAVA_HOME so that it may find core
classes and tools. It has to discover a suitable JAVA_HOME (or verify
the one specified by the user) first, and then run
${JAVA_HOME}/bin/java. And now that you refreshed my memory regarding
the logic in javavmwrapper (already exporting JAVA_HOME to its
children), I think we may already greatly simplify or improve this
launcher shell script.

Thanks for reminding me of this, Greg. :)

BTW, Sam sent me a patch using the method #2 (overridable JAVA_HOME) I
suggested so I guess we are fine for now regarding NetRexx/JAVA_HOME.

OTOH, looking at '/usr/local/bin/maven':

  TOOLS_JAR="${JAVA_HOME}/lib/tools.jar"

So that's another use case for the %%JAVA_HOME%% feature you suggested.

Now we have to decide whether it's worth implementing the feature. But I
think we need to find out exactly the ports that would benefit from such
feature...

Herve