Thanks. Sadly the patch doesn't quite work. I attached the log file.
I tried a small fix to the patch, ppc-ldb2-beo.diff. Sadly that just gets me back
to the corrupted image problem as before.
On Wed, Apr 18, 2012 at 11:00:28AM -0400, Paul Khuong wrote:
> On 2012-04-18, at 4:21 AM, Christophe Rhodes wrote:
> > I don't have access to any PPC systems at all. The patch needs at a
> > minimum a test case or several, to try to stop this coming in again. I
> > think it's also a little bit too conservative on signed-num and
> > unsigned-num primitive type arguments; shouldn't there be
> > signed=>unsigned and unsigned=>unsigned variants of %%LDB, with size and
> > position ranges (integer 1 32) and (integer 0 31) respectively,
> > retaining the constraint that size + position <= 32.
> Sounds right, although that kind of chance always make me afraid we're introducing more opportunities for bad representation selection.
> > Also, Paul Khuong was threatening on IRC to submit a cleaned-up version
> > of my patch -- Paul, what's your plan?
> The attached patch should be the exact same, with less copy paste. Unfortunately, I don't have access to a PPC box either (: