[RFC] target/m68k: multilib issues

[Finn Thain]

On Thu, 18 Nov 2010, Thorsten Glaser wrote:

> Finn Thain dixit:
> 
> >I believe it is supported by newlib.
> [?]
> >Dropping cpu32 from m68k-linux doesn't solve the gfortran problem for 
> >uclinux.
> 
> True. Yes, this should be fixed on the gcc level, although it won?t 
> reflect in Debian because I only need to build multiarch/lib for things 
> that are actually usable there.

Right, glibc uses CAS instructions which are not implemented on CPU32. But 
the larger question of m68k-uclinux and m68k-linux ABI compatibility is 
one I can't answer, though I had the impression that m68k-uclinux had 
adopted the m68k-linux ABI.

> So, no input from me on the gcc-level problem, sorry. (Does newlib use 
> m68k-unknown-linux-gnu as well? I thought that was for (e)glibc only.)

I don't know. In my tests I only built uclinux and linux cross-compilers. 
Hence, the patch I sent only disables fidoa mulitlibs, and doesn't affect 
cpu32 or any other multlibs.

Finn

> 
> bye,
> //mirabilos
>