cross-compiling

Joachim Beckers joachim.beckersNOSPAM at belgacom.net
Thu Jan 22 11:02:17 PST 2004


James Robertson wrote:
> Ryan.Oliver at pha.com.au wrote:
>[..] 
> IIUC, this conversation and the work being performed is in conjunction 
> with the cross-compiling hint which is as you mention about faking a 
> compiler into thinking it is an earlier x86 proc.
 >
> Having nALFS be able to do *real* cross-compiling whereby you are using 
> an x86 to compile binaries for say an Alpha is not in scope with right 
> now AFAIK.  I really think that this could be accomplished with a 
> specially built profile to handle all the hacks.
IMO, this could be accomplished, but it will be a tough job and it will take a 
long time.

Important question:
How do you test if a "real" cross-compile profile works?
You'll always need someone who has two computers from a different architecture.
There's no way you'll be able to test all possible combinations of 
host/target-architecture.

Besides, the only reason I can see for one to crosscompile an LFS, is to speed 
up the build because it would take too long on the target computer, being an old 
x86, apple/macintosh or other slow machine.
(Are there other old architectures one would crosscompile for?)

Why would you want to build an LFS on your (let's say) 1GHz athlon for your 1GHz 
powerpc?

Conclusion:
IMO, making a profile to crosscompile LFS for a different sub-architecture will 
be difficult enough.
Making one to do *real* crosscompiling will be too difficult (at least for me:-)

comments welcome

Joachim Beckers

-- 

-----------------------------------------------------------
Please remove NOSPAM in my e-mail adress to reply.
-----------------------------------------------------------




More information about the alfs-discuss mailing list