Re: [MiNT] NOCFLAGS $(MODEL)

[Alan Hourihane <alanh@fairlite.co.uk>]

On Wed, 2011-03-23 at 00:12 +0100, Vincent Rivière wrote:
> Helmut Karlowski wrote:
> > I agree, but at least you know what files need to be patched.
> 
> I always know which files need to be patched, because when they are compiled 
> with the normal CPU flags for the current configuration, the assembler fails 
> with "unsupported instruction" or something like this.
> 
> > You say they compiled without error?
> 
> Yes. And as a test, I have removed all those NOCFLAGS for C files and 
> compiled all the kernels variants, there was absolutely no compile error 
> (remember I have just patched the .c/.h files for ColdFire). However those 
> kernels should be tested.
> 
> > Why have they been added?
> 
> This is my question.
> 
> > I know I got compiler-errors for some XaAES-files for coldfire and just
> > imitated what was done in the kernel.
> 
> I hope those NOCFLAGS have not been introduced for compiling that old 
> FreeMiNT for FireTOS. Let's see the Makefiles.obj CVS history...
> Horror!
> 
> Revision : 1.5
> Date : 2006/12/3 19:41:41
> Author : 'landemarre'
> State : 'Exp'
> Lines : +36 0
> Description :
> Some files can't compil with -mcfv4e option, need choose other options (some 
> instruction will be emulate)
> 
> So it seems the speed of *all* the kernels has been sacrificed to allow 
> building an experimental and partial ColdFire kernel. And that change lasted :-(
> 
> Since I'm currently patching the FreeMiNT sources for ColdFire using the 
> right way, this is another reason to get rid of that NOCFLAGS abomination.

Yes. Go for it.

Also, all the modules such as the XDD's, XIF's etc are compiled for
68000. This isn't right. All the modules should be in their own
processor directory (I have mentioned this before).

We really need to fix the Makefiles, including a "make install" option.

Alan.