[OE-core] [oe-core][PATCH 11/13] libdrm: improve packaging

[Richard Purdie]

On Fri, 2011-09-30 at 09:27 +0200, Martin Jansa wrote:
> Signed-off-by: Martin Jansa <Martin.Jansa at gmail.com>
> ---
>  meta/recipes-graphics/drm/libdrm.inc       |    7 ++++---
>  meta/recipes-graphics/drm/libdrm_2.4.26.bb |    2 +-
>  meta/recipes-graphics/drm/libdrm_git.bb    |    2 +-
>  3 files changed, 6 insertions(+), 5 deletions(-)
> 
> diff --git a/meta/recipes-graphics/drm/libdrm.inc b/meta/recipes-graphics/drm/libdrm.inc
> index 1ad9dc3..20d7f1f 100644
> --- a/meta/recipes-graphics/drm/libdrm.inc
> +++ b/meta/recipes-graphics/drm/libdrm.inc
> @@ -15,7 +15,8 @@ DEPENDS += " libpciaccess"
>  
>  inherit autotools pkgconfig
>  
> -PACKAGES =+ "libdrm-tests"
> -FILES_libdrm-tests = "${bindir}/dr* ${bindir}/mode*"
> +PACKAGES =+ "${PN}-tests ${PN}-drivers ${PN}-kms"
> +FILES_${PN}-tests = "${bindir}/dr* ${bindir}/mode*"
> +FILES_${PN}-drivers = "${libdir}/libdrm_*.so.*"
> +FILES_${PN}-kms = "${libdir}/libkms*.so.*"
>  
> -LEAD_SONAME = "libdrm.so"

I want to make it *really* clear that this breaks anything depending on
libdrm for oe-core users.

This is exactly why I think the whole PR bump thing is stupid, braindead
and simply wrong since the current fix is to go and find anything using
libdrm and bump those PR values too.

Of course we could add a "LIBDRM_INC_PR" in the core and bump that when
libdrm changes.

The trouble is following this path to its logical conclusion, we
basically turn our automated build system into a set of manual changes
where we compute dependency chains by hand.

Sorry to labour this point but I do want people to understand why we
need to take some pain in the short term and improve what we're doing...

I will take the patch since I value getting meta-oe cleaned up and I'll
add in a patch bumping PR for xserver/mesa but I'm not happy.

Cheers,

Richard