[Buildroot] [PATCH 1/2] infra: improve dependency check in virtual packages.

Mike Zick minimod at morethan.org
Mon Feb 24 18:58:05 UTC 2014


On Mon, 24 Feb 2014 19:29:51 +0100
"Yann E. MORIN" <yann.morin.1998 at free.fr> wrote:

> Eric, All,
> 
> On 2014-02-24 10:50 +0100, Eric Le Bihan spake thusly:
> > The current version of dependency check for virtual package <foo>
> > defines FOO_CONFIGURE_CMDS to print an error message if the
> > dependencies are not met.
> > 
> > This patch updates all the virtual packages to use the GNU Make
> > control function $(error text...) instead.
> > 
> > This makes the error happen at the beginning of the build, with a
> > clearer message.
> > 
> > Signed-off-by: Eric Le Bihan <eric.le.bihan.dev at free.fr>
> 
> Reviewed-by: "Yann E. MORIN" <yann.morin.1998 at free.fr>
> 
> Somewhat-unrelated to this change, see a comment below...
> 
> > ---
> >  package/luainterpreter/luainterpreter.mk |    6 ++++++
> >  package/opengl/libegl/libegl.mk          |    7 +++----
> >  package/opengl/libgles/libgles.mk        |    7 +++----
> >  package/opengl/libopenmax/libopenmax.mk  |    7 +++----
> >  package/opengl/libopenvg/libopenvg.mk    |    7 +++----
> >  package/powervr/powervr.mk               |    7 +++----
> >  6 files changed, 21 insertions(+), 20 deletions(-)
> > 
> > diff --git a/package/luainterpreter/luainterpreter.mk
> > b/package/luainterpreter/luainterpreter.mk index c37d621..5443477
> > 100644 --- a/package/luainterpreter/luainterpreter.mk
> > +++ b/package/luainterpreter/luainterpreter.mk
> > @@ -9,4 +9,10 @@ LUAINTERPRETER_DEPENDENCIES = $(call
> > qstrip,$(BR2_PACKAGE_PROVIDES_LUA_INTERPRET 
> >  LUAINTERPRETER_ABIVER = $(call
> > qstrip,$(BR2_PACKAGE_LUAINTERPRETER_ABI_VERSION)) 
> > +ifeq ($(BR2_PACKAGE_HAS_LUA_INTERPRETER),y)
> > +ifeq ($(LUAINTERPRETER_DEPENDENCIES),)
> > +$(error No lua interpreter implementation selected. Configuration
> > error.) +endif
> > +endif
> > +
> >  $(eval $(generic-package))
> > diff --git a/package/opengl/libegl/libegl.mk
> > b/package/opengl/libegl/libegl.mk index b2b74f1..3311e50 100644
> > --- a/package/opengl/libegl/libegl.mk
> > +++ b/package/opengl/libegl/libegl.mk
> > @@ -7,11 +7,10 @@
> >  LIBEGL_SOURCE =
> >  LIBEGL_DEPENDENCIES = $(call
> > qstrip,$(BR2_PACKAGE_PROVIDES_OPENGL_EGL)) 
> > +ifeq ($(BR2_PACKAGE_HAS_OPENGL_EGL),y)
> >  ifeq ($(LIBEGL_DEPENDENCIES),)
> > -define LIBEGL_CONFIGURE_CMDS
> > -	echo "No libEGL implementation selected. Configuration
> > error."
> > -	exit 1
> > -endef
> > +$(error No libEGL implementation selected. Configuration error.)
> > +endif
> >  endif
> >  
> >  $(eval $(generic-package))
> 
> I don't know about you folks, but I'm beginning to see a pattern
> here...
> 
> If all of our BR2_PACKAGE_HAS_FOOBAR options would match the
> corresponding FOOBAR_DEPENDENCIES, then we could use something like
> (absolutely untested, only pure random thoughts):
> 
>     $(eval $(call virtual-package,FooBar))
> 
> whith 'virtual-package' something like:
> 
>     define virtual-package
>     ifeq ($(BR2_PACKAGE_HAS_$(call upper,$(1))),y)
>     ifeq ($($(call upper,$(1))_DEPENDENCIES),)
>     $(error No $(1) implementation selected. Configuration error.)
>     endif
>     endif
>     endef
> 
> But then we would need:
>     s/BR2_PACKAGE_HAS_OPENGL_EGL/BR2_PACKAGE_HAS_LIBEGL/
>     s/BR2_PACKAGE_HAS_OPENGL_GLES/BR2_PACKAGE_HAS_LIBGLES/
>     s/BR2_PACKAGE_HAS_OPENMAX/BR2_PACKAGE_HAS_LIBOPENMAX/
>     s/BR2_PACKAGE_HAS_OPENVG/BR2_PACKAGE_HAS_LIBOPENVG/
> 
> Lua interpreter and PowerVR already match this.
> 
> We could even go further, so as to only require a call to
> 'virtual-package' and not to 'generic-package', with some
> carefully-crafted Makefile constructs (but we would loose the
> capitalisation of eg. lbEGL, since the package is named 'libegl', all
> in lower-case). 'virtual-package' could even override
> FOOBAR_SOURCE="" before calling generic-package behind the hood. This
> would make writing virtual packages even easier.
> 
> Thought?
> 

Sounds good to me.
(The first suggestion, without the lost capitalization.)

Mike

> Regards,
> Yann E. MORIN.
> 



More information about the buildroot mailing list