Kevin Fisher <kgf at golden.net> is widely believed to have written:

> This is still using an unoptimized 2.9a VM (due to ARM GCC troubles).  
>
Some work on the NetWinder equivalnet problem finally got done the other
day; it looks like the optimizer generates a particular floating point
instruction that is not being suported properly on the latest NW kernel.
It looks like the dame problem afflicts the iPaq, so I'm suspicous of
the overall toolchain. I have passed an explanation of the evidence so
far to the appropriate (I think) people. Curiously, the older NW kernel
has no problem with the instruction and merrily runs the optimised VM.
Go figure.

tim
-- 
Tim Rowledge, tim at sumeru.stanford.edu, http://sumeru.stanford.edu/tim
Useful Latin Phrases:- Nihil curo de ista tua stulta superstitione = I'm not interested in your dopey religious cult.