Re: Once more cross compilation

[BALATON Zoltan]

On Thu, 22 Jul 2004, Nicola Pero wrote:
> > What if I compile gnustep-make with HOST=x86, TARGET=arm and try to
> > crosscompile another package with it. Then I should compile the other
> > package with HOST=arm. Does this conflict with the usage of this variable
> > in gnustep-make?
>
> (this question is about compiling a crosscompiler using a crosscompiling
> gnustep-make)

No it isn't. There was no TARGET involved in the compilation of the other
package. The question is the same if you crosscompile any package (say
GNUMail which has no TARGET) with a crosscompiling gnustep-make.

> By the way, there would be a relationship between your compiler's
> variables and gnustep-make's ones ... Your compiler's build platform will
> be GNUSTEP_HOST_* (that is because the system where you run gnustep-make
> is the system where you build your compiler), your compiler's host
> platform will be GNUSTEP_TARGET_* (that is because the system where you
> run the compiler is the system where you run applications created with
> gnustep-make), and your compiler's target platform ... well you have to
> configure that into the compiler - this has nothing to do with
> gnustep-make.

Exactly, (and you see, TARGET is not involved). Now the question is:
should each package really use different BUILD/HOST/TARGET varibles or can
gnustep-make be made smart enough to support this in a way to allow
crosscompiling all packages (including gnustep-make) in the same way (e.g.
by saving the HOST/TARGET variables when compiling gnustep make and using
them under different, non-conflicting names internally later)?

Regards,
BALATON Zoltan