Re: raw1394_request user/kernel space size differences

[Ben C. <bco...@de...>]

> Actually, I have implemented a simple but effective fix now.  Kernel
> side doesn't change, libraw1394 casts pointers to kernel pointer width
> if neccessary.  For now configure requires --enable-32-64 on the
> command line, there is no automated way to recognize a splitted
> environment.

This is very wrong. For example, I should be able to compile a 32bit sparc
binary that runs on sun4cdm (32bit kernel arch) and sun4u (64bit kernel
arch) just like any other binary does. This makes the resulting binary
specific to the kernel, and this I wont have that kind of portability.
This might be ok on a per user basis, but when libraw1394 hits the
distributions (and it will be in Debian soon I suspect), it will play hell
with packaging. ia64, ppc64 an mips64 will have the same problem. The
32bit binaries should be portable, and the "hackery" should be in the
kernel so it is invisible to userspace.

Take modutils for example. It is only compiled as 32bit, and the resulting
binary can handle 32bit and 64bit modules.

Ben

-- 
 -----------=======-=-======-=========-----------=====------------=-=------
/  Ben Collins  --  ...on that fantastic voyage...  --  Debian GNU/Linux   \
`  bco...@de...  --  bco...@op...  --  bco...@li...  '
 `---=========------=======-------------=-=-----=-===-======-------=--=---'