[Buildroot] svn commit: trunk/buildroot/toolchain/dependencies

[Bernhard Fischer]

On Tue, Aug 21, 2007 at 10:58:40PM +0200, Ulf Samuelsson wrote:
>tis 2007-08-21 klockan 22:25 +0200 skrev Bernhard Fischer:
>> On Tue, Aug 21, 2007 at 10:12:59PM +0200, Ulf Samuelsson wrote:
>> >tis 2007-08-21 klockan 13:14 -0700 skrev aldot at uclibc.org:
>> >> Author: aldot
>> >> Date: 2007-08-21 13:14:23 -0700 (Tue, 21 Aug 2007)
>> >> New Revision: 19631
>> >> 
>> >> Log:
>> >> - make sure that sstrip is available early enough
>> >> 
>> >> 
>> >> Modified:
>> >>    trunk/buildroot/toolchain/dependencies/dependencies.mk
>> >> 
>> >> 
>> >> Changeset:
>> >> Modified: trunk/buildroot/toolchain/dependencies/dependencies.mk
>> >> ===================================================================
>> >> --- trunk/buildroot/toolchain/dependencies/dependencies.mk	2007-08-21 19:45:17 UTC (rev 19630)
>> >> +++ trunk/buildroot/toolchain/dependencies/dependencies.mk	2007-08-21 20:14:23 UTC (rev 19631)
>> >> @@ -5,7 +5,11 @@
>> >>  #
>> >>  ######################################################################
>> >>  
>> >> -dependencies: host-sed host-lzma
>> >> +ifeq ($(BR2_STRIP_sstrip),y)
>> >> +# XXX: this is a little bit ugly, yep.
>> >> +MAYBE_SSTRIP_HOST:=sstrip_host
>> >> +endif
>> >> +dependencies: host-sed host-lzma $(MAYBE_SSTRIP_HOST)
>> >>  	@HOSTCC="$(firstword $(HOSTCC))" MAKE="$(MAKE)" \
>> >>  		HOST_SED_DIR="$(HOST_SED_DIR)" \
>> >>  		$(TOPDIR)/toolchain/dependencies/dependencies.sh
>> >> 
>> >
>> >Why not add host-sed/host-lzma (conditionally) to TARGETS early enough?
>> >I tried this, and this seemed to work, but I never got to commit that.
>> >
>> >ifeq ($(BR2_STRIP_sstrip),y)
>> >TARGETS+=sstrip_host
>> >endif
>> >
>> >in toolchain/Makefile.in
>> >
>> >Should do wonders.
>> 
>> I expect this not to work.
>> Try the usual testcase for this kind of stuff:
>> set BR2_JLEVEL=1
>> $ make clean
>> $ make -j 64
>> (or 128)
>> and see it go up in flames if you don't have proper deps in certain
>> sequence points..
>> 
>> See?
>
>Oh, Oh... Yes!
>
>It would be nice to have a 64-128 CPU build server making the problem
> dead obvious :-)

Since you're most likely I/O bound anyway (you most likely don't build
in a ramdisk or something that is vaguely fast for IO) there's no point
in letting your one to eight cores be idle. No need for a 512 CPU box to
make that obvious.
>
>Whatabout:
>
>HOSTTOOLS:=host-sed host-lzma
>ifeq ($(BR2_STRIP_sstrip),y)
>HOSTTOOLS+=sstrip_host
>endif

Except that "HOSTTOOLS" is a too generic name, yes.

MAYBE_HOST_PREREQ is what i would have used (since lzma should also be
conditional). Care to do this?

>dependencies:  $(HOSTTOOLS)

TIA,