> : Interesting! I didn't know it could be done in user mode for MIPS, I
> : it had to be done no later than during boot (I seem to remember a pin on
> : chip for this?).
> That's why Ralf keeps talking about the bi-endian kernel. There is a
> bit in the PSW that says "I'm running in the opposite byte order now"
> that just needs to be toggled, and a few other kernel<-> changes.
> Well, and some library changes (htons is a noop on BE, but requires
> code on LE). You wouldn't believe the bugs you get when you are
> running in one endian and using numbers from the other endian once in
> a while...
GNU libc is _very_ in clean with respect to byte order problems. I found
a single byte order problem yet which was crosscompiling related. If
you tried to compile the lib for a big endian machine the crosscompiler
host's byteorder was actually used which of course fails on a Intel box ...
The big problem is in the kernel.