Re: libada cross-compilation

[Nick Burrett <nick at sqrt dot co dot uk>]

Richard Earnshaw wrote:
> On Fri, 2005-02-25 at 09:47, Nick Burrett wrote:
>
> > I found the following patch was required to allow for the successful 
> > building of the gnattools for a cross-compiler.  Without this patch, we 
> > would be using a mix of the built cross-compiler and the host's gnattool 
> > -set.
> >
> > Built on i686-pc-linux-gnu and cross-compiled to arm-unknown-elf.
> >
> > If this is OK, can somebody commit it as I don't have write-access.
> >
> >
> > <date> Nick Burrett  <nick@sqrt.co.uk>
> >
> > 	* Makefile.in (gnattools-cross): Use native CC.
> >
> > ______________________________________________________________________
> > --- /home/nick/riscos-aof/masters/gcc-4.0/libada/Makefile.in	2004-05-30 09:51:07.000000000 +0100
> > +++ libada/Makefile.in	2005-02-25 09:42:01.142391312 +0000
> > @@ -164,6 +164,7 @@
> > gnattools-cross: gnatlib
> > 	$(MAKE)  -C $(GCC_DIR)/ada $(FLAGS_TO_PASS) $(ADA_FLAGS_TO_PASS) \
> > 	  ADA_INCLUDES="-I$(RTS_DIR)../adainclude -I$(RTS_DIR)" \
> > +	  CC="gcc" \
> > 	  GNATMAKE="gnatmake" \
> > 	  GNATBIND="gnatbind" \
> > 	  GNATLINK="gnatlink" \
>
>
> This can't be right.  It's never correct to hard-bind CC to a specific
> compiler name (with the exception of the compiler we've just built).
>
> Perhaps you mean $(CC_FOR_BUILD)?

I agree with what you're saying, but we already hard-bind GNATMAKE, 
GNATBIND and GNATLINK.  We have to pick up the CC that is in the same 
path as GNATBIND otherwise we get build-inconsistencies.

If we were to use CC_FOR_BUILD, then there should be a 
GNATBIND_FOR_BUILD, GNATMAKE_FOR_BUILD and GNATLINK_FOR_BUILD.

I don't like it.

Nick.