[RFC] add --with-build-for-cc option

[Robert Schuster]

Hi all,

Am 21.06.2010 15:47, schrieb Xerxes Rånby:
> Hi Robert!
> 
> ...
> diff -r 1e95c787d40b ports/hotspot/make/linux/makefiles/zeroshark.make
> 
> --- a/ports/hotspot/make/linux/makefiles/zeroshark.make	Tue Jun 15 18:34:45 2010 +0100
> +++ b/ports/hotspot/make/linux/makefiles/zeroshark.make	Fri Jun 18 13:30:12 2010 +0200
> @@ -35,6 +35,12 @@
>  #XXX and the updated calling convention for deopt (PR icedtea/484)
>  #CFLAGS += -DHOTSPOT_ASM
>  
> +ifeq ($(CROSS_COMPILATION), true)
> +  C_COMPILE_FOR_MKBC = $(CC_FOR_BUILD)
> +else
> +  C_COMPILE_FOR_MKBC = $(C_COMPILE)
> +endif
> +
>  %.o: %.S
>  	@echo Assembling $<
>  	$(QUIETLY) $(REMOVE_TARGET)
> @@ -53,7 +59,7 @@
>  
>  mkbc:	$(GAMMADIR)/tools/mkbc.c
>  	@echo Compiling mkbc tool
> -	$(CC_COMPILE) -o $@ $< $(COMPILE_DONE)
> +	$(C_COMPILE_FOR_MKBC) -o $@ $< $(COMPILE_DONE)
>  
>  mkoffsets:	asm_helper.cpp
>  	@echo Compiling offset generator
> 
> ...
> 
> I do not see why we need to conditionalize the zeroshark.make code at all,
> I think it would be better and cleaner to always use the passed
> $(CC_FOR_BUILD) when compiling mkbc.c
Yep, I can understand that but there is a point for having both:

a) CC_COMPILE is derived from the variables in that OpenJDK makefile
that sets up gcc. AFAIK it carries optimization flags and can be
modified by changing CFLAGS for example. CC_FOR_BUILD instead is a
variable that only contains the value for the compiler and does not
react to changes to the CFLAGS flags. I wanted to not touch the normal
build in any way (so nobody could say that it broke their build on
Ubuntu/Debian or Fedora ;) )

That said, I see that my patch changes the mkbc compilation from
CC_COMPILE to C_COMPILE. That confusion must somehow slipped in because
I originally prepared the patch against IcedTea6 1.7.3. Before
committing I would fix that.

b) As said in the mail before I have some different plans for dealing
with mkoffsets. At the moment we 'solve' the situation by (please don't
beat me) by running it through qemu's ARM emulation. In order to support
that transparently (= zero effect for non-cross builders) I would like
to use the "ifeq ($(CROSS_COMPILATION)" section again.

Thanks for having a look at the patch!

Regards,
Robert

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 262 bytes
Desc: OpenPGP digital signature
Url : http://mail.openjdk.java.net/pipermail/distro-pkg-dev/attachments/20100621/5ffba705/signature.asc