RE: Moving system properties to gnu.classpath.*

[Jeroen Frijters]

Michael Koch wrote:
> Personally I'm against both types of such hacks. The current solution 
> probably duplicates some code but it easy to understand, needs no 
> hacks and extra support from the VMs.

I don't think the current solution is easy to understand. The library
initialization bootstrapping is quite complex because many classes
require access to the system properties and currently that forces
java.lang.System to be touched (which in turn constructs
System.in/out/err and that sets off more static initializers).

This is not just a theoretical issue for me. IKVM is very different from
most VMs in this respect. I don't have a well defined place to
initialize the library, any class can be the first to be used by client
code.

In most cases this works fine, except for some scenarios where
gnu.java.io.EncodingManager is initialized before java.lang.System, then
it fails because of a cyclic dependency.

Regards,
Jeroen