This is the mail archive of the
mailing list for the GCC project.
Re: [RFC] target/m68k: multilib issues
- From: Maxim Kuvyrkov <maxim at codesourcery dot com>
- To: Finn Thain <fthain at telegraphics dot com dot au>
- Cc: gcc-patches at gcc dot gnu dot org, Thorsten Glaser <tg at mirbsd dot org>
- Date: Fri, 19 Nov 2010 14:02:27 +0300
- Subject: Re: [RFC] target/m68k: multilib issues
- References: <alpine.OSX.2.00.1011180049460.1202@localhost>
On Nov 18, 2010, at 1:50 AM, Finn Thain wrote:
> BTW, on the topic of multilib issues, has anyone managed to build
> --with-arch=cf --enable-multilib? It doesn't build for me. Is this
> configuration supported?
Yes, it is supported. This is how all CodeSourcery's ColdFire toolchains are built.
> 2010-11-17 Finn Thain <email@example.com>
> * config/m68k/m68k-devices.def (FL_MMU, FL_UCLINUX): Remove.
> * gcc/config/m68k/m68k.h (FL_MMU, FL_UCLINUX): Remove.
> * config/m68k/t-linux (FL_MMU): Remove.
> (M68K_MLIB_CPU): Specify multilibs explicitly. Drop fidoa.
> * config/m68k/t-uclinux (FL_MMU, FL_UCLINUX): Remove.
> (M68K_MLIB_CPU): Specify multilibs explicitly.
I don't think specifying multilibs explicitly is a good approach for M68K port. It seems cleaner to annotate the device descriptions with flags specifying which target triplet a particular device supports.
Now, I do agree that the flag names are not very descriptive of their effect, that is due to historic reasons. "FL_MMU" is equivalent to "supports GNU/Linux", and "!FL_MMU || FL_UCLINUX" is equivalent to "supports uClinux", and bare-metal target is naturally supported for any device. I think it would be a nice cleanup to transform these into self-descriptive FL_LINUX and FL_UCLINUX flags.