[pylucene-dev] JCC 1.8 - double arrays as arguments in java methods

[Andi Vajda]

On Sat, 19 Apr 2008, Bill Janssen wrote:

>> I also just released JCC 1.9 to cheeseshop [2] with this and many other
>> fixes since release 1.8:

> I was able to build 1.9 (shared) for Windows XP, Python 2.5, after
> installing setuptools.  I've now got bdist packages for that
> configuration; I believe you will need setuptools to install them on a
> target machine, though.  Would it be useful for me to put them
> somewhere so others could download them?

As I've done before for other user-built binaries, if you make them 
available somewhere for the world to download, I'll put up a link to the 
message announcing them on the PyLucene homepage. You can put any 
information specific to your binaries in that message.

> I believe the references to the location of Java (and the particular
> micro-release of Java being used) are still hard-coded into the JCC
> libraries, so perhaps not.  It would be easy (I've got the code) to
> teach JCC to look in the registry to find Java, but when you build
> PyLucene with JCC, is the location of libjava hard-coded into that
> PyLucene binary?

On Linux and Solaris, the location of runtime libraries is hardcoded into 
the binaries produced if you've chosen to do so by using the -R or -rpath 
linker flag. This is recommended as having better security than setting 
LD_LIBRARY_PATH. Whether you choose to hardcode the location or not, you can 
override it with LD_LIBRARY_PATH (I think ?).

On Mac OS X, the location of Java is determined by the OS. Java is a 
pre-installed system framework.

On Windows, I don't think this can be hardcoded at all (I think ?) and the 
Java binaries are found by the PATH. You have shown before that PATH can be 
changed at runtime _before_ these libraries are loaded. I don't hardcode any 
such PATH setting into JCC anywhere. It's up to you to do this whichever way 
works best for your situation (registry trick or otherwise).

Andi..