Re: (native build & other issues - Was: xvfb not built on vax...)

[David Brownlee <abs%NetBSD.org@localhost>]

On Wed, 4 Feb 2009, Brad Parker wrote:

> >         b) Groff chokes building certain .ps files - I thought Brad had
> >            tracked that down to a hypot issue, what ever happened with
> >            that?
>
> mmm.  I found the problem but was unsure that to do.  My fix was to
> replace the hypot() call with a "local_hypot()" which did the right
> thing but did not cause an exception.
>
> as I recall instead of doing something rational like SIGFPE the low level
> hypot code signaled an error with an illegal instruction.
>
> I guess I needed more guidance.  I can go back at it.  I was doing
> native builds on a simulated 11/780 at the time and they completed
> over night, so it was not hard to set up & test.

        I would be inclined to start with generating a test case
        - copy regress/lib/libm/round to regress/lib/libm/hypot
        and update to trigger the issue. That gives us a reference
        point, and something to trigger if the bug shows up again
        in future.

        Then its probably putting your fixed version in lib/libm/arch/vax
        and getting the system to use it :)

> >         Can people please take a look at http://wiki.netbsd.se/vax
>
> very nice!
>
> I want to help Dave McGuire get netbsd going on an MVII some time in the
> future.  I'm up the the 11/730, which makes me happy.  I thank I have
> the parts to put together an MVII in my back room.

        Most excellent!

--
                David/absolute       -- www.NetBSD.org: No hype required --