Hi Jocke,
The new patches cause a build failure in libm when building for classic
powerpc.
I've attached a patch to fix this,
Best regards,
Steve
On Sat, 2007-03-31 at 15:29 +0200, Joakim Tjernlund wrote:
> Merged, please check if I missed something.
>
> Jocke
>
> > -Original Message-
> > From: Steve Papacharalambous [mailto:[EMAIL PROTECTED]
> > Sent: den 20 mars 2007 08:55
> > To: Joakim Tjernlund
> > Cc: uclibc@uclibc.org
> > Subject: RE: e500 libm patches.
> >
> > Hi Jocke,
> >
> > Thanks for getting back to me about the e500 uClibc libm patches.
> > Please see in line answers,
> >
> > Best regards,
> >
> > Steve
> >
> > On Sat, 2007-03-17 at 17:58 +0100, Joakim Tjernlund wrote:
> > > Hi Steve
> > >
> > > I started to look at your math patches again :)
> > >
> > > > -Original Message-
> > > > From: Steve Papacharalambous [mailto:[EMAIL PROTECTED]
> > > > Sent: den 23 februari 2007 14:22
> > > > To: uclibc@uclibc.org
> > > > Cc: Joakim Tjernlund
> > > > Subject: e500 libm patches.
> > > >
> > > > Hi,
> > > >
> > > > I've done an initial port of the e500 libm fpu code to uclibc, and
> > > > modified them to include the comments from Joakim Tjernlund.
> > > >
> > > > These patches can be downloaded from the website:
> > > > http://www.bitshrine.org/gpp/
> > > >
> > > > The patches move all the existing code in libm/powerpc to
> > > > libm/powerpc/classic and put the e500 code in libm/e500/fpu
> > > >
> > > > Patch #1:
> > > >
> > http://www.bitshrine.org/gpp/uclibc-20070119-e500-libm-classic.patch
> > > >
> > > > This patch move all the existing files in libm/powerpc to
> > > > libm/powerpc/classic.
> > > >
> > > > Patch #2:
> > > > http://www.bitshrine.org/gpp/uclibc-20070216-e500-libm-e500.patch
> > > >
> > > > This patch add e500 code to libm/powerpc/e500 and
> > > > libm/powerpc/e500/fpu
> > > >
> > > > Patch #3:
> > > >
> > http://www.bitshrine.org/gpp/uclibc-20070216-e500-libm-string.patch
> > > >
> > > > This patch add the e500 string conversion functions.
> > >
> > > There are a lot of string funs(ato..., strto..) that nobody
> > > uses. Who is the user? gcc?
> >
> > These will be used by the printf and scanf extensions for the e500 to
> > add long double support. I haven't done the patches to the printf and
> > scanf functions needed to use these functions yet as I was
> > waiting until
> > all the main e500 code had been merged.
> >
> > However if you want to drop long double support for now we can just
> > ignore the patch: uclibc-20070216-e500-libm-string.patch and we can
> > look at this later on.
> >
> > > Is it possible to build a non math enabled uClibc system?
> > >
> >
> > Not sure what you mean here? If you're asking whether uClibc can be
> > configured and built for e500 and soft float then the answer is yes.
> >
> > > The file machine-ieeefp.h contains a lot of non powerpc
> > > code, why is this file placed here?
> >
> > Mainly because this it is only used by the string functions and it was
> > in the local directory in the glibc port, so I left it in the
> > same place
> > for the uClibc port.
> >
> > However, if you want to ignore support for long double for the time
> > being and not apply the libm-string patch mentioned above, then
> > machine-ieeefp.h will not exist.
> >
> > >
> > > There also appers to be long double support in here
> > > and uClibc does not support long double yet and when
> > > it does there will probably be wrapper funs that
> > > maps the long double to just double to keep the
> > > size down. Lets drop long double for now.
> >
> > The removal of the libm-string patch should remove all the long double
> > support code.
> >
> > >
> > > fprrest.c fprsave.c are empty files
> > >
> >
> > Deleted.
> >
> > > There are a few:
> > > versioned_symbol (libm, __feclearexcept, feclearexcept, GLIBC_2_2);
> > > that's is very glibc only ATM.
> > >
> >
> > Removed these lines.
> >
> > >
> &