Re: Soft-float and newlib

[msokolov at ivan dot Harhan dot ORG (Michael Sokolov)]

Walter Christian <wolti@sil.at> wrote:

> Yes. I just found the msoft-float library version of newlib in the
> source-tree. You were right and it was build because it was part of the
> multilib variant.

Ahmm, it isn't in the source tree, it's *installed* in the appropriate
subdirectory under your ${prefix}/m68k-coff/lib. It is automatically used when
you link with gcc and give the gcc link command the same options you gave the
gcc compile commands, including -msoft-float.

> > 000050B4 E9C2               DC.W        $E9C2
>
> The function where the exception occured is __fixdfsi. As you said that
> soft-float is part of gcc does that mean that my gcc compiler doesn't work
> as it should ?

Hmm, so __fixdfsi has an opcode in it that's illegal for CPU32, right? gcc's
software floating point emulation library should support CPU32, as it's
supposed to support even the plain 68000, of which CPU32 is a superset. If it's
supposed to work but doesn't, it would be a bug in gcc. I can't troubleshoot
this right now, you'll have to wait until I get settled in California. You
might want to bug me in a few weeks.

OTOH, it might not be a bug in gcc, but a user error. :-) For the m68k family
the CPU selection options have a profound effect not only on the compiler, but
also on the assembler. Even though the assembly code for gcc's soft-float
library is probably correct for all CPUs, for different multilib variants it
gets assembled with different options, and the assembler might very well be
generating different opcodes. Are you sure that you are compiling everything
with gcc -mcpu32 -msoft-float and doing the final link with gcc -mcpu32
-msoft-float?

--
Michael Sokolov		Harhan Engineering Laboratory
Public Service Agent	International Free Computing Task Force
			International Engineering and Science Task Force
			615 N GOOD LATIMER EXPY STE #4
			DALLAS TX 75204-5852 USA

Phone: +1-214-824-7693 (Harhan Eng Lab office)
E-mail: msokolov@ivan.Harhan.ORG (ARPA TCP/SMTP) (UUCP coming soon)