libgcc c++ support bits moved

[Jason Merrill]

>>>>> "Todd" == Todd Vierling <tv@wasabisystems.com> writes:

> On Mon, 9 Oct 2000, Richard Henderson wrote:
> : All the C++ support bits have now been moved out of libgcc.

> : The removal of the C++ bits from libgcc also allows some of the build
> : infrastructure to be simplified a smidgin.

> Please document this in the texi files, and also fix the texi files to
> indicate that the "gcc" frontend program mustn't be used to link C++
> programs, even those not using the standard library's facilities.

> In the past, for quite some time (and there are programs depending on this
> fact), C++ code not making use of "C++ includes" (aka standard library in
> today's terms) could be compiled and linked with the "gcc" program.

This has *always* been the case, before this change.  I know I have
always used 'gcc' for linking simple C++ programs.

> As of this change, this will no longer be possible without
> explicitly saying -fno-exceptions -fno-rtti to gcc.

Yep.  And not even then, if they use new and delete.

My strong preference would be to leave the stuff in libgcc somehow,
but I accept that popular opinion is against me (and that I'm weighing
in far too late to get a vote).

However, I still think it's important to keep the freestanding support
library separate from the hosted library, since the latter has a bunch
of startup code that should not be imposed on folks that don't use the
functionality.

This would seem to call for a third library, perhaps "libgpsup"
(libg++ is already taken).  It should probably go in another toplevel
directory.

Thoughts?

Jason