Re: [PATCH] [BZ#2749] Allow powerpc32 to build --without-fp

[Steven Munroe <munroesj at us dot ibm dot com>]

Joseph S. Myers wrote:
>On Mon, 24 Jul 2006, Daniel Jacobowitz wrote:
>
>  
>>On Wed, Jun 14, 2006 at 01:54:42PM -0500, Steven Munroe wrote:
>>    
>>>This patch combined with the other patches from Bugzilla #2749 allows
>>>powerpc32 to build. The Subdirs pulls in the soft-fp directory from
>>>libc, The Implies pulls in the ports powerpc/soft-fp directory which
>>>includes sfp-machine.h
>>>
>>>The get/set/swapcontext changes overide the common implimentation in
>>>libc to avoid using hardware FP instructions.
>>>      
>>What's the status of this patch?  Who needs to approve it or commit it?
>>I appear to have write access to all of ports, so I could do it myself,
>>but I figured I ought to check with Roland since I haven't touched
>>anything outside of the ARM/MIPS ports before.
>>
>>I don't really want to be responsible for all of ports, but I don't
>>want patches to linger indefinitely either.
>>    
>
>In addition, as I mentioned at the GCC Summit, could someone revert
>
>2006-01-27  Roland McGrath  <roland@redhat.com>
>
>        * sysdeps/powerpc/nofpu/Makefile [$(subdir) = soft-fp]
>        (sysdep_routines): Add $(gcc-quad-routines) here.
>and remove the *tf* functions from sysdeps/powerpc/nofpu/Versions?  The 
>*tf* functions, for IEEE quad long double, should not be used on PowerPC, 
>and since they haven't appeared in a release with functional --without-fp 
>PowerPC support, it is still safe to remove them from libc.
>
>  
My team is working on debugging the remaining make check failures which
is slow going.  Also I believe that Khem Ral has built powerpc
-without-fp so I would like some feadback on any issues/fixes he may
have found.

We have some addional patches that need to be reviewed and posted at
least to the bugzilla. This would include the changes that Joseph
mentioned. But I understand that some non-linux system use the IEEEquad.
So I would leave the IEEE quad code in place but update
sysdeps/powerpc/Makefile to not build them.

It does seem to be a backlog of patches for both libc and ports. I
assume the core maintainers are busy with other work. Perhaps we can
help by taking more of the work load for ports by having additional
platform commit-with-approval maintainers.