[omniORB] Re: omniORB and egcs-1.1.1 for Win32

[Sai-Lai Lo]

Hello Steven,

I don't know of anyone trying omniORB2 with egcs-1.1.1 for windows. As the
source already compiles with egcs on other platforms, I guess there won't
be much problem there.

However, you may encounter some difficulties in building DLLs and importing
DLLs. The way MSVC++ does this at the moment is diabolical. Let me know if
you need some explanation on how we currently build the DLLs.

The thread issue is interesting. Surely a pthread implementation would go
someway to address the issues of doing Current object.


Sai-Lai



>>>>> Brenneis, Steven W writes:

> Hello Sai-Lai, 
> I have downloaded cygwin-b20 and egcs-1.1.1 for Windows.
> The compiler is very complete and the build system works flawlessly.  For
> a number of reasons, including some of the issues I raised previously
> (COS Security, CORBA::Current, etc.), the egcs compiler has become very
> attractive.  I have puttered around a little attempting to build omniORB
> using this compiler but time constraints prevent me from devoting as much
> attention to it as I would like.  I got as far as building omniidl2,
> which works well.  I started to build omnithread, but hit a stumbling
> block with the underlying implementation. Has anyone begun work on
> producing an egcs-Win32 implementation of omniORB?  I do not want to
> duplicate anyone's efforts.

> There has been much discussion in the cygwin mailing lists surrounding
> their standard library implementation (the weakest part of the egcs
> system, IMHO).  Some of this discussion involves a pthread
> implementation, which I feel sure they are on the verge of producing.  I
> even went as far as to try importing the FreeBSD version of pthread, but
> weaknesses in the g++ runtime library stopped me.  I believe POSIX
> threads availability would go far to solving the problems of
> CORBA::Current and would produce a stronger implementation of omniORB in
> general.  What do you think?