Brain-dead compilers [was: [omniORB] Re: compiling omniORB3 o n NT]

Chung, David David.Chung@USPTO.GOV
Wed, 27 Oct 1999 15:20:29 -0400


	I also hope that omniDynamic and omniORB3 can be cleanly separated
out.  From looking at various omniORB related email messages, it seems that
the original design was to allow omniORB3 to link with apps without
omniDynamic anyways (unless omniDynamic specific stuff were used).

> -----Original Message-----
> From:	visschb@rjrt.com [SMTP:visschb@rjrt.com]
> Sent:	Wednesday, October 27, 1999 2:19 PM
> To:	omniorb-list@uk.research.att.com
> Cc:	djr@uk.research.att.com
> Subject:	Brain-dead compilers [was: [omniORB] Re: compiling omniORB3
> on NT]
> 
> Hello omniORBers,
> 
> David Riddoch wrote:
> > Yes -- you should be able to link to libomniORB* without linking to
> > libomniDynamic*.  However, this doesn't seem to work at the mo'.
> 
> This is a problem with DEC CXX version 5.6 on VMS also (even at 2.x.y;
> Note: this is not a problem with CXX 6.2).  This significantly
> complicates creating shareable images on VMS (esecially on VAX). 
> Basically, if the compiler sees just about any external references the
> linker is going to want to have a definition.  Even if they are not
> used.  I.e., to really seperate omniORBx from omniDynamicx, we would
> need to prevent the compiler from seeing TypeCode, Any, etc. (other than
> forward declarations) unless something "dynamic" is being compiled. 
> While I think this is just plain dumb (on the part of the compiler), is
> this enough of a problem on enough platforms to be worth the effort to
> somehow seperate this?
> 
> Bruce Visscher