This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Patch for Bug#12955
Jason Merrill <jason@redhat.com> writes:
> On Mon, 10 Nov 2003 07:52:22 -0800, Michael Eager <eager@mvista.com> wrote:
>
> > In two places I added #if 0's which surround code which tests for
> > ROUND_TOWARDS_ZERO. There are no targets which implement this option, so
> > it isn't easy to check if the code is correct or not. I'm pretty sure
> > that the low-order bit proxy which I added is not compatible with it
> > for the non-ROUND_TOWARDS_ZERO case. I left the code in simply in
> > case it may be needed if ROUND_TOWARDS_ZERO is ever used.
>
> That makes sense, but you should add a comment to that effect. Or leave
> the code enabled, but add an abort if it is actually used so that someone
> has to fix it.
fp-bit.c is linked into the user's program. I don't think an abort is
appropriate for an unlikely floating point rounding case. Better
would be #error in the ROUND_TOWARDS_ZERO case.
Ian