On Wednesday 12 September 2001  8:41 pm, Antti Kantee wrote:
> Ah, this one. I really didn't bother fixing the problem, since I don't
> claim to understand how the source tree in structured at the moment, and
> what it's supposed to look like in the future (if someone would point me
> to a plan, I might be able to do some grunt work here, hint hint).
> Instead I just hacked around the problem.

As requested some info on what I believe are current plans:
Split out the real platform independant stuff into the arm and arm32 dirs.
Currently hpcarm has a lot of duplication,  it would be good to see if pulled 
back in, however there are some stumbling blocks to this. one obvious one 
being that hpcarm has generic soft interrupts, arm32 doesn't.  We need to 
pull this into the common arm32 dir.

It would also be cool if hpcarm also used the arm/sa11x0 dir eventually.

The aim is to make the arm kernels look the same to userland, this will allow 
the same ARM ELF binaries to run across all hardware.

> memory_disc_size is declared somewhere under arm32/rc7500, which is not
> compiled into the hpcarm iPAQ kernel. So I just simply added the
> declaration to some random file. Removing the keyword `extern' in
> arm32_machdep.c should be enough to get the kernel linked.

I believe it's something to do with some extra stuff in the 
arm32/dev/md_hooks.c file.  Inparticular the ability to load the memory disk 
image from a floppy disk, this is possibly no longer needed/used as kernels 
now have the disk image embedded in them

> ps. someone should actually fix this

So many things so little time.  I really should start examining our 
infrastructure for the arm kernels and the arm installs etc.

Chris