32bit/64bit datatype issues

[Richard Franks]

We had some major headaches with this - mostly because legacy code
written for 32bit architectures tends to make silly assumptions that
pointers can be cast to integers. But there also a number of tricky
cases where it wasn't immediately obvious that the datatype
discrepancy was the root cause.

As it becomes harder to find 32bit desktop processors, this is going
to become increasingly significant, especially for services or
applications which communicate between the Neo1973 and a 64bit
home/work PC.

I noticed that the GIMP project has its own typedefs (guint, guint8,
gint16, etc).. so I was wondering if OpenMoko will be supplying its
own definitions too? Going into the future, it would mean that there
is a level of protection against having to go back through reams and
reams of code and finding obscure errors that could have been avoided
very easily.

On a more present-day note, if it had native conversions for big/small
endianness, that would be really really nice, too.

Richard