darwin cross compiling breakage

[Ian Lance Taylor]

Mike Stump <mrs@apple.com> writes:

> On Jun 12, 2006, at 12:25 PM, Andrew Pinski wrote:
> > A little more, the default CROSS for a cross compiler defines
> > CROSS_COMPILE and
> > NON_SAME_CROSS_COMPILE while Darwin only defines CROSS_COMPILE.
> 
> And so you want all of the existing:
> 
> #ifdef CROSS_COMPILE
> 
> to be changed to:
> 
> #if defined(CROSS_COMPILE) && defined(NON_SAME_CROSS_COMPILE)
> 
> save for the two in i386.h?  That's doable, but Geoff's already said:

Presumably he means to change every

#ifdef CROSS_COMPILE

to

#ifdef NON_SAME_CROSS_COMPILE

Let me note that I count some 34 uses of CROSS_COMPILE in gcc.  Of
those, only 8 are in machine independent code.  3 more are code that
might be used in Darwin.  So we're not talking about some sort of
heroic effort here.

The machine independent uses of CROSS_COMPILE can mostly be
characterized as "don't use native tools" and "don't use native header
files and libraries."  So we could introduce something like
NO_NATIVE_TOOLS.  Then we could change every machine independent
instance of CROSS_COMPILE (all 8 of them) to check NO_NATIVE_TOOLS
instead.  We could change other cases, where appropriate, at our
leisure.

Then a normal cross would define both CROSS_COMPILE and
NO_NATIVE_TOOLS.  A Darwin -> Darwin cross would define CROSS_COMPILE
but not NO_NATIVE_TOOLS.

I don't think this would be confusing, nor do I think it would cause
an ongoing maintenance burden.

Ian