NetBSD-4 softfloat (was: Apache on Mac68k)

Hi,

When running with build.sh I didn't have a mk.conf, rather, 
I just set MKSOFTFLOAT=YES in the environment first.

Therefore I don't know.  OTOH, this is the first crash is
the -Werror flag which is normally a good idea.
On Mon, Dec 18, 2006 at 04:58:55PM +0900, Kazuyuki Inanaga
wrote:
> Hi,
> 
> I've been trying to build.sh release for FPU-less 68k
Macs
> on cross-compiling, and I think it has got a
"success".
> 
> base machine: NetBSD/i386(4.99.6)
> source: -current (4.99.6)
> 
> Case-1: /etc/mk.conf
>         MKSOFTFLOAT=yes
>         COPTS+=-msoft-float
> 
> It got "Error Stop" both on i386 and
Quadra800(4.99.4) at;
> 
> >#  compile GENERIC/bus_space.o
> >[..]
> >cc1: warnings being treated as errors
> >/usr/src/sys/arch/mac68k/mac68k/bus_space.c: In
function  
> >'mac68k_bsr2_gen':
> >/usr/src/sys/arch/mac68k/mac68k/bus_space.c:418:
warning: statement  
> >with no effect
> 
> I'm sure this -current source sets has no problem for
normal
> compiling. I have no idea to avoid this Error Stop.
> 
> 
> Case-2: /etc/mk.conf
>         MKSOFTFLOAT=yes
> 
> Without "COPTS+=-msoft-float", it has
completed build.sh release
> including -x.

Which, unless the makefiles have changed alot, the
-msoft-float 
flag should have been passsed to the c compiler.
> 
> 
> Then I installed those binary sets to LC III and
External HDD
> for LC475 and PB520.
> NO error is found from the installation to the starting
xdm.
> (LC III gets panic at reboot, it's same as
"with-FPU" machine.
> And I don't know how to run ColorX on SOFTFLOAT.)
> 
> No more than 10 hours operation on LC475, it seems to
work
> "perfect".  NetBSD is great! Many thanks to
the people who
> are developing NetBSD.
> 
> 
> I'm wondering why does "COPTS+=-msoft-float"
spit "Error Stop"?
> 
> >gcc-4.1.1                2006-05-06               
GCC(1)
> >
> >M680x0 Options
> >
> >-msoft-float
> >Generate output containing library calls for
floating point.
> >Warning: the requisite libraries are not available
for all m68k
> >targets.  Normally the facilities of the machine's
usual C compiler
> >are used, but this can't be done directly in
cross-compilation.
> >You must make your own arrangements to provide
suitable library
> >functions for cross-compilation.  The embedded
targets m68k-*-aout
> >and m68k-*-coff do provide software floating point
support.
> To my regret, I don't understand this well. I have to
learn more,
> especially in English? ;)
> 
> 
> New(?) xdm is running on LC475 now. It is very nice.
> 
> On 2006/12/09, at 0:00, Bruce O'Neel wrote:
> >
> >I've very partial to tvtwm.  Set with no window
borders and
> >almost no color.  I find it very relaxing and quick
to work
> >with.  Others in my office seem to hate it.  Works
with
> >m68k systems on up.
> >
> >My .tvtwmrc file is:
> I copied it. Thanks a lot!
> >I find it very relaxing and quick to work with.
> I can guess so.
> >Others in my office seem to hate it.
> 
> 
> 
> I have one more question.
> Is it required "MKSOFTFLOAT=yes" or anything
for making packages?
> I forgot to set it before starting make, but it seems
OK.
> 
> Thanks,
> Kazu Inanaga
> 
> 

For building packages you need to have -msoft-float passed
to the 
C compiler.  Otherwise if the package does FP calculations
it
will use the FP instructions and things won't work right.

cheers

bruce


-- 
edoneelsdf.lonestar.org
SDF Public Access UNIX System - http://sdf.lonestar.org

Hello folks,

How are you doing in this beautiful and lovely season?

I've been busy with work and fixing my mail system more
sucure
because my last posts were not delivered to this list and
Michael R. Zucca. I'm sorry.

Thanks a lot to Bruce O'Neel for covering it.
Bruce O'Neel wrote:
> Hi,
> 
> When running with build.sh I didn't have a mk.conf,
rather, 
> I just set MKSOFTFLOAT=YES in the environment first.
(snip)
  Kazuyuki Inanaga wrote:
  >> I have one more question.
  >> Is it required "MKSOFTFLOAT=yes" or
anything for making packages?
  >> I forgot to set it before starting make, but it
seems OK.

> For building packages you need to have -msoft-float
passed to the 
> C compiler.  Otherwise if the package does FP
calculations it
> will use the FP instructions and things won't work
right.

I understand. Thanks, again.
I'm still not clear about some points, I think "traial
and error"
is a good way for me now. 

Regards,
Kazu Inanaga