[openbeos] Re: Misc. problems building Haiku on Linux

On 2005-11-19 at 01:00:00 [+0100], Ingo Weinhold wrote:
> On Sun, 13 Nov 2005, Oliver Tappe wrote:
[ 8< ]
> > Something at the back of my head is telling me that there was some 
> > problem with binutils > 2.15, but currently I can't remember what it 
> > was (maybe it was an incompatible interface change in ehopt.c [of which 
> > we use an old, forward-ported version], maybe it was something else). I 
> > guess the only way to find out is to try the update >:o/
> 
> Mmh, considering how much time you spent to "fine-tune" the binutils-gcc 
> combo, I don't feel so motivated to experiment with newer binutils in 
> this context. I think I'll rather move this binutils version + gcc 2.95.3 
> into a subdirectory "legacy" and import a current binutils version 
> alongside with gcc 4, renaming the "gcc4" directory to just "gcc".

Yep, that sounds sensible and safe, because "legacy" is just what that 
stuff really is... >;o)

> > Having looked into the mentioned problem just now, I suppose the 
> > problem could be fixed by moving the definition of struct relax_type 
> > from tc.h to as.h (where, according to a comment in tc.h, it used to 
> > live at some time in the past). I think the theory mentioned in the 
> > comment at tc.h:28 is no longer true (with gcc4), so we might just as 
> > well move the definition back into as.h.
> >
> > Could someone that has encountered the mentioned build issue with gcc4 
> > please refer to the enclosed patch and tell me if that fixes it?
> 
> I've just updated my gcc under Linux to 4.0.2 and could reproduce the 
> problem and also verified that the patch fixes it. Both binutils and gcc 
> 2.95.3 build just fine here now (SuSE 9.2). Shall I check your patch in, 
> or do you want to do it?

No, just go ahead please and check it in.

> My next steps on the gcc 4 front will be to clone the i586-pc-beos 
> configurations of binutils and gcc to i586-pc-haiku, create 
> powerpc-???-haiku (any idea about the middle part?) configurations and 
> get cross compilation for those working...

What kind of hardware would it actually run on? Macs, I assume and probably 
some other, more obscure stuff, right (like IBMs >;o)? Maybe we should just 
go with 'apple' or 'unknown'?

cheers,
        Oliver

Other related posts: