RE: Patch for PowerPC + uClibc + softfloat

["Andy Johnson" <ajohnson at aecno dot com>]

Everyone,

> On Friday 16 January 2009 21:15:35 Andy Johnson wrote:
> > I have put together the 360-powerpc-uclibc-softfloat.patch
> > for GCC 4.3.2 by shamelessly lifting a patch from a thread
> > on the gcc-patches mailing list.  Without going into the
> > ugly details, this patch was originally written for the
> > 4.4.0 trunk (but later withdrawn), but I have backported
> > it to 4.3.2 so crosstool-ng can build for the PowerPC using
> > uClibc and softfloat.
> 
> Why was it withdrawn? Was it superseded by another patch?

This fix was withdrawn not because it didn't work, but
because they agreed that it should be fixed a different
way.  So as far as I know, this is the only fix for us
for now.  Even if uClibc gets fixed, this patch would
still be necessary for anyone who needs to use an older
version (or the version current today), like me.
 
> As you present it, this patch is supposed to fix a mis-compilation
> for PowerPC, but it touches only generic files, and no PowerPC files.

True.  Basically all this patch does is add libm to the
configure tests and makefile for the PowerPC + uClibc
combination so that an unresolved reference to copysignl
can be found in libm instead.  No source is changed.

> That patch won't go in. It's wrong, as per the thread you pointed to.

The problem won't ultimately be fixed in the compiler build
script, but rather in uClibc.  I haven't tried any of the
uClibc snapshots to see if this problem has been fixed yet.

In this case it is possible to fix this particular
problem more than one way.  For now though, this is our
only fix.

Andy

--
For unsubscribe information see http://sourceware.org/lists.html#faq