[osmosis-dev] Plugin Bug

[marcus.wolschon at googlemail.com]

On Mon, 2 Feb 2009 12:12:51 +0100, Jochen Topf <jochen at remote.org> wrote:
> Hi!
> 
> I tested the plugin problem further. It works on this machine:
>     java version "1.6.0_07"
>     Java(TM) SE Runtime Environment (build 1.6.0_07-b06)
>     Java HotSpot(TM) Client VM (build 10.0-b23, mixed mode, sharing)
> 
> But not on this:
>     java version "1.6.0_10"
>     Java(TM) SE Runtime Environment (build 1.6.0_10-b33)
>     Java HotSpot(TM) 64-Bit Server VM (build 11.0-b15, mixed mode)
> 
> Maybe its something to do with 32 vs 64 bit.
> 
> I tested this without Marcus' plugins stuff, just used the -p option and
> the same problem happens. As Tiles at Home uses -p and thats run on many
> machines, I suspect that this might have also been broken by the new
> plugin framework.

Hello,

just to shed some light here.
The old plugin-system should still work, but I did not test it
as I had no plugins for it.
The new one just puts a new way to load instances of
com.bretth.osmosis.core.plugin.PluginLoader
in front of the old one (I did not want to change too many things).

I can rule out that Java 1.6.0_10 is the reason as I am using _10
here on a 32bit-system with no issues. (SSL-handling changed in _10
but that does not affect us.)

What exactly did you try with the "-p" -option?
Did whatever you tried work in the 32bit-machine?
Could you send me your old-style plugin, so I can
check that the old plugin-api is indeed still working
and if not fix it?

I may try to download a 64bit linux-distribution and
install it in a 64bit virtual machine emulated on a 32bit-one.
However this will take a loooong time to do on my laptop.
As your NullPointerException is inside the JavaPluginFramework
I may find a simple testcase to send them and file a bug
in their tracker. (I would never have thought something with
no native IO-code to fail on a 64bit machine but work on a
32bit one.)

Marcus