Larry V. Streepy, Jr. <[EMAIL PROTECTED]> wrote:
> 1. When processing an interface (like a Home interface), the _Stub
> class name will be the same as the interface. E.g., VersionHome
> becomes _VersionHome_Stub. No tie class is created.
Like I said somewhere at the beginning of this thread (more than a
week ago), I'm not really familiar with RMI in the IIOP world, I'm
just judging from documentation and some experiments.
I didn't even know you could rmic an interface when using -iiop (you
can't do that for plain RMI) - this means we'd need some additional
checks anyway, oh my.
> Does everyone agree that this seems to capture the name generation
> accurately?
Yes.
> Stefan, where are you at on this?
A first pass without reflection is already in CVS, it does the right
thing for the _Tie classes (if you rmic the implementation). Our
discussion has brought up a couple of things (and made me read some
more documentation) that still need to be folded in.
rmic will need some more modifications - we need to handle -idl case
properly (don't even try to find generated files) and allow verify and
iiop/idl to work together (Ant simply says you cannot rmic an
interface).
> SHould I wait for your patches, or try to implement the above
> myself?
I'll do so, thanks.
Stefan
