Re: corba idl4 and the hurd ?

[Niels Möller]

Etienne Robillard <address@hidden> writes:

> 1. I cannot compile idl4 because I'm getting 'virtual
> memory exhausted' messages while compiling
> serverloop.cc ... Any ideas on what could be the cause
> of this ? 

No idea, I managed to compile it some week ago.

> 2. My question is for the implementation of the Hurd
> root_server : can't we use optimized stub_code to
> simplify/enhance sync IPCs procedures that involves
> servers process ? 

I don't think so. I think we should either

  * use idl4 as is,

  * or help improve idl4,

  * or write our own stub generator

I think it's quite hard (perhaps impossible) to write optimal ipc stub
code in C, using the official L4 interface, and the current gcc. And
you don't want to hand code assembler for each and every interface.

This applies to all Hurd server (or rather, Hurd interfaces, the stub
code is specific to an interface, not to a particular server or
client), I don't see a reason why the root server should be important
or different in this context.

Regards,
/Niels