Re: Patches for the Hurd on OSF Mach/PowerPC

[Peter Bruin]

Marco Gerards <metgerards@student.han.nl> wrote:

> I had a really quick look at your patches, especially the ext2fs
> patch. I have a little (perhaps stupid) question about this patch. I
> usually use htons ntos, etc. for handling endianess. Was there a
> reason not to use these functions? I'm really happy to see those
> improvements for endianess :). I still wonder if other parts of the
> Hurd won't have these problems...

The ext2 file system uses little-endian order, whereas htons etc. convert from
native to network (big-endian) byte-order.  So using htons would give you
exactly the wrong result.  (Actually, I didn't even think of using htons, since
I looked at the Linux code for ext2fs and copied the le**_to_cpu functions from
there.)

I haven't tried the other filesystems yet, but isofs and ufs might have the same
problem.  I don't know about fatfs; I see Marcus's implementation already uses
endian-dependent functions for reading and writing, but I haven't tested it yet.

> Could you please write ChangeLog entries. I know this is boring work,
> but that makes the code easier to read and it is a requirement for
> submition AFAIK.

Hmm, I knew I forgot something... 8-)  Thanks, I'll do that!

Peter