Re: [PATCH 0/3]: Add math support for non LDBL_EQ_DBL architecture

[Joel Sherrill <joel@rtems.org>]

[-- Attachment #1: Type: text/plain, Size: 1896 bytes --]

On Thu, Apr 6, 2023 at 4:00 AM Corinna Vinschen <vinschen@redhat.com> wrote:

> On Apr  5 10:44, Joel Sherrill wrote:
> > On Wed, Apr 5, 2023 at 4:23 AM Corinna Vinschen <vinschen@redhat.com>
> wrote:
> > > I noticed that you exclude Cygwin from the new code, which makes sense
> > > as long as we provide our own long double math taken from Mingw-w64.
> >
> > What needs to happen for Cygwin/Mingw_w64? Honestly, we didn't discuss
> > that at all.
>
> Not sure I understand the question.  After the patch, Cygwin should
> just still work.  Mingw-w64 isn't supported by newlib.  What do you
> mean?
>

I was just worried there was something special we needed to account for
that we didn't. I'm glad we didn't miss anything there. No worries. :)

>
> > > However, there's something not quite right.  When trying to build I get
> > > symbol conflicts for the fdim{f,l} and scalbln{f,l} symbols in the link
> > > stage (paths shortend for readability):
> > > [...]
> >
> > We chatted and those will be fixed in the next round.
>
> Thanks.
>
> > Anything else you think needs tidying up?
>
> Well, yes.  The patch introduces a new header <sys/endian.h>.  Cygwin
> already maintains an <endian.h> header, as on Linux.
>
> The way we install header files is that first, the newlib headers are
> installed, and then, second, Cygwin headers are installed, thus
> overriding newlib headers with Cygwin headers as required.  We have to
> maintain <endian.h> for userspace compatibility, but after the
> installation user will find two different versions of endian.h,
> one in /usr/include, one in /usr/include/sys.
>
> Would it make a lot of problems to move <sys/endian.h> to <endian.h>?
>

Probably not. Just a matter of the final <endian.h> being a proper union
of the two. It's just software after all.

--joel

>
>
> Thanks,
> Corinna
>
>
>