going forward

[Jakub Jelinek]

On Fri, Aug 17, 2001 at 02:39:53PM -0700, Ulrich Drepper wrote:
> Roland McGrath <roland@frob.com> writes:
> 
> > Enough has been done already, that I would hope we could get this fully
> > resolved by a week from today.
> 
> I'm not so optimistic.  So far all I have seen are results from people
> running completely new environments and individual results of running
> older programs.  Nobody put down which combinations have been tested.

I was doing some testing, particularly:
main C++ program compiled with 2.96-RH and 3.0.1 throwing exceptions through
C library -fexceptions compiled with 2.96-RH and 3.0.1 an all tests passed.
If main program was compiled with 2.96-RH, I had to use -lc explicitely on
the command line, since otherwise __frame_state_for and registry/unregistry
from libstdc++ is used, but this is a distribution compatibility thingie
which is not related to glibc - I'll just patch all compatibility libstdc++
libraries so that they don't have the register/unregister/__frame_state_for
stuff at all.

As 2.96-RH's __frame_state_for is now binary compatible with egcs 1.1.2 and
2.95.x, I don't think there is a stressing need to do the testing with all
compilers, but it of course won't hurt.

But surely it needs more testing, so I'll do some testing myself after the
weekend and ask other people to do some testing too once I build rpms out of
it.

	Jakub