svn commit: r307761 - in head: . contrib/binutils/bfd contrib/gcc/config/rs6000 gnu/lib/libgcc gnu/lib/libgomp gnu/usr.bin/binutils gnu/usr.bin/cc gnu/usr.bin/cc/cc_tools gnu/usr.bin/cc/include gnu...

[Konstantin Belousov]

On Sat, Oct 22, 2016 at 03:20:17PM +0300, Konstantin Belousov wrote:
> On Sat, Oct 22, 2016 at 01:57:15AM +0000, Justin Hibbits wrote:
> > Author: jhibbits
> > Date: Sat Oct 22 01:57:15 2016
> > New Revision: 307761
> > URL: https://svnweb.freebsd.org/changeset/base/307761
> > 
> > Log:
> >   Create a new MACHINE_ARCH for Freescale PowerPC e500v2
> 
> > Added:
> >   head/lib/libc/powerpcspe/sys/ptrace.S   (contents, props changed)
> 
> This is not needed, libc provides MI implementation in libc/sys/ptrace.c.

Why the decision to copy whole lib/libc/powerpc to lib/libc/powerpcspe
was used to create the architecture ? I did a primitive-science
overview, and got total ~2300 lines in lib/libc/powerpcspe, while
$ diff -Nrua lib/libc/powerpc lib/libc/powerpcspe | diffstat -p0 | tail -1
 16 files changed, 240 insertions(+), 174 deletions(-)

It seems to be possible to only create several new files, like
*setjmp.S and fp*.c, and reuse existing stock files from libc/powerpc
for anything else.

I am the one who does most of the sweeping changes over libc, and
I am somewhat saddened by addition of almost identical copy of the
existing architecture files which need to be maintained together with
the original.