Re: [uClinux-dev] Re: [PATCH] Add m68k-uclinux target

[Bernardo Innocenti <bernie at develer dot com>]

Christoph Hellwig wrote:

> > [...]
> > Unlike the official, 2.95.x based toolchain, GCC 3.4 will
> > stop abusing the m68k-elf target and use its own target
> > triplet instead (m68k-unknown-uclinux).
> > [...]
>
> Are you sure it should really be m68k-unknown-uclinux and not
> m68k-unknown-linux-uclibc?

You're right, but I don't want t-linux to be used in GCC.
There are too many changes there to be worth sharing.
Perhaps m68k-unknown-uclinux-uclibc would be an even
better name for it.

> AFAICS the libstdc++ changes are
> for uclibc and not for a no-mmu kernel.

Yes, you're right.

> And given that 'uclinux' in the kernel context (vs the distro
> context) is nothing but a Linux kernel with a single ifdef
> not set (or a patched kernel with a single ifdef set
> for < 2.6) a different target name looks very strange for me.

You're forgetting that the userland changes quite a lot in ways
that matter to the toolchain: different linker configuration,
the FLAT binary format, id-based shared libraries...

Besides, the missing CONFIG_MMU in uClinux makes a lot of things
much different. Missing mmap() and fork() make the system no
longer POSIX compliant!

Configure scripts for most applications will need to know about
that. Since you will be cross-compiling, you can't even run
tests for that. The only way to tell whether you can use these
functions is by looking at the target triplet.


> And the libcstdc++ changes look right for uclibc on targets with
> a mmu aswell, dito for !m68k nommu targets.  

Yes, the -uclibc postfix is definitely something to add, also
because you could also be using uClinux with glibc or uC-libc.

--
 // Bernardo Innocenti - Develer S.r.l., R&D dept.
\X/  http://www.develer.com/

Please don't send Word attachments - http://www.gnu.org/philosophy/no-word-attachments.html