--=-=-=
Content-Transfer-Encoding: quoted-printable

Oleg Polyanski <Oleg.Polianski@clear.co.nz> writes:

 Oops, forgot to add that until the recent binutils (2.11.x series) had been
 integrated into -current the exception handling didn't work because the
 compiler was not using the `-fsjlj-exceptions' option by default, it had to
 be specified manually. If you have the plain installation of 1.5.x/i386
 series somewhere around and few spare minutes, could you please check it
 (sorry, I don't have any PC available for experiments right now). Just in
 case, tiny test program is attached.

>  I've found my PR #8465 dated with 21 September of 1999 which was closed =
in
>  August of 2000. For some weird reasons I've never received this reply wi=
th
>  notification and kept using `-fsjlj-exception' until recent binutils have
>  been integrated into -current. Anyway, `State-changed-why' claims that
>=20
> State-Changed-Why:=20=20
> This was solved by using sjlj exceptions on all elf ports (8 June 2000)=
=1B:=1B[3=20
>=20
>  So, it seems that my information is correct.
>=20
> > : > author referred).  a.out machines such as this use sjlj exceptions =
by
> > : > default.
> > :
> > :  .... no, they don't. Tested and proved.
> >=20
> > Is there an open PR bug report on this?
> >=20
> > I realize that 1.4.x is now considered `old', but I'm the person who
> > imported and configured the toolchain used in 1.4.x.  The g++ used on
> > 1.4.x/i386 should have been using sjlj exceptions, as that was specific=
ally
> > the way I set it.

Take care,
O


--=-=-=
Content-Type: application/octet-stream
Content-Disposition: attachment; filename=exc.C

class A { };

int
main (void)
{
	try {
		throw A ();
	}
	catch (A &e) {
		return 1;
	}
	catch (...) {
		return 2;
	}

	return 0;
}

--=-=-=--