PPC64 Compiler bug !!

[linas at austin.ibm.com]

On Thu, Jun 12, 2003 at 03:11:39PM -0500, Peter Bergner wrote:
> linas at austin.ibm.com wrote:
> > OK, that's a sane answer.  I first saw this when compiling a custom
> > kernel module; I see that the standard linux kernel uses -msoft-flost
> > in its standard Makefile, and I will be tracking down this custom code
> > shortly.
>
> Using -msoft-float when compiling the kernel and/or kernel modules is a
> _requirement_!  We've hit too many user data integrity errors due to
> third party modules not being compiled with -msoft-float, which is why
> I changed the kernel src to force a panic if we take a FP Unavailable
> exception within the kernel.

That's fine, that is in fact how we tripped over this thing.  I have
no complaints, it just sort of came as an unexpected surprise.  I've
mucked with gcc internals and .md in the past, and even with that
background, it never occured to me that this was a 'feature' not a
'bug'.   Seems that not too many people know about this.

Most people probably never trip on this, since they suck in the
standard kernel makefile; I was dealing with a non-standard setup.

--linas

** Sent via the linuxppc64-dev mail list. See http://lists.linuxppc.org/