On Mon, 1 Nov 1999, Christos Ricudis wrote:

> I have a FPU-less LC II (1.4.1 and current tested) which core dumps with
> "illegal instruction" errors when trying to compile the kernel. A simple
> program using libm also produces the same error. Seems that the system is
> working as no FPU emulation was present (it is). 

If it were a simple case of missing FPU emulation, I would expect the 
"illegal instruction" errors to be 100% predictable and happen in the same
exact place and circumstance with complete reproducibility.

If repeated attempts to compile code fail randomly, at different places
each time, then I would instead suspect buggy FPU emulation.

Random and unpredicable failures (illigal instruction and segmentation
fault) are what I see on my Powerbook 540c (but then, nobody promised me
working 68LC040 support).

I've been able to get the compilations to at least finish by using the
following (bourne) shell construct:

$ while ! make
> do
> done

For each compilation failure, the shell will dutifully restart
it.  Eventually, after many failures along the way, the compilation
finishes.  Perhaps this approach will help you as well, though your
68030 is likely to have different problems than my 68LC040.

> And what about the programs that use floating poing but do not link with 
> libm? (as, for one, is not linked with libm)

Perhaps these can be recompiled with gcc's -msoft-float option, but I
haven't tried it.
-- 
Tod McQuillin