[meta-intel] emgd-driver-dev and libgl-mesa-dev packaging conflict

Tom Zanussi tom.zanussi at intel.com
Fri Apr 19 13:51:09 PDT 2013


On Fri, 2013-04-19 at 13:46 -0700, Darren Hart wrote:
> 
> On 04/19/2013 12:23 PM, Tom Zanussi wrote:
> > On Thu, 2013-04-18 at 22:37 -0500, Kamble, Nitin A wrote:
> >>>>>
> >>>>> With IPK packaging I am able to reproduce the issue. I think this can
> >>>>> be easily avoided by reverting this commit
> >>>>>
> >>>>> commit 5e314278e4cf37b7fbb2946f4f42a3236071e131
> >>>>> Author: Laurentiu Palcu <laurentiu.palcu at intel.com>
> >>>>> Date:   Wed Mar 13 13:05:48 2013 +0200
> >>>>>
> >>>>>     emgd-driver-bin: add pkgconfig files
> >>>>>
> >>>>>     Add pkgconfig files needed by other projects (for example newer libva
> >>>>>     packages look for egl.pc file).
> >>>>>
> >>>>>     Pkgconfig files added by this patch:
> >>>>>      * egl.pc
> >>>>>      * gl.pc
> >>>>>      * gles_cm.pc
> >>>>>      * glesv2.pc
> >>>>>
> >>>>>     Signed-off-by: Laurentiu Palcu <laurentiu.palcu at intel.com>
> >>>>>     Signed-off-by: Tom Zanussi <tom.zanussi at linux.intel.com>
> >>>>>
> >>>>>
> >>>>>
> >>>>> This commit was needed for newer libva, now we have pinned EMGD
> >>> based
> >>>>> BSPs to older libva, so this commit is not really needed.
> >>>>
> >>>> I verified that this indeed fixes the issue.
> >>>
> >>> It more like hides the issue right?
> >> No, it is not really hiding. This commit adds the conflicting files, which are not needed anymore.
> >>
> >>>
> >>> The problem seems to be a continued conflict between emgd and mesa
> >>> supplying the same files.
> >>>
> >>> Seems to me we need something like the anonymous python in the meta-
> >>> intel mesa bbappend from:
> >>>
> >>> commit 1a5ab65544830dd571a368b9d9f6ec782ffc94ac
> >>> Author: Nitin A Kamble <nitin.a.kamble at intel.com>
> >>> Date:   Tue Oct 30 14:38:34 2012 -0700
> >>>
> >>>     mesa-dri.bbappend: avoid conflict with emgd-driver-bin
> >>>
> >>> Right?
> >> That can be another way to fix the problem. But the issue is not only with mesa-dri, I was seeing conflicts with other recipes too as seen below.
> >>
> >> | Collected errors:
> >> |  * check_data_file_clashes: Package libgl-mesa-dev wants to install file /srv/home/nitin/build-test-bsps/build-emenlow/tmp/work/emenlow-poky-linux/core-image-sato-sdk/1.0-r0/rootfs/usr/lib/pkgconfig/gl.pc
> >> | 	But that file is already provided by package  * emgd-driver-dev
> >> |  * opkg_install_cmd: Cannot install package qt4-x11-free-dev.
> >> |  * check_data_file_clashes: Package libgl-mesa-dev wants to install file /srv/home/nitin/build-test-bsps/build-emenlow/tmp/work/emenlow-poky-linux/core-image-sato-sdk/1.0-r0/rootfs/usr/lib/pkgconfig/gl.pc
> >> | 	But that file is already provided by package  * emgd-driver-dev
> >> |  * opkg_install_cmd: Cannot install package mesa-demos-dev.
> >> |  * check_data_file_clashes: Package libgl-mesa-dev wants to install file /srv/home/nitin/build-test-bsps/build-emenlow/tmp/work/emenlow-poky-linux/core-image-sato-sdk/1.0-r0/rootfs/usr/lib/pkgconfig/gl.pc
> >> | 	But that file is already provided by package  * emgd-driver-dev
> >> |  * opkg_install_cmd: Cannot install package libva-egl-dev.
> >> |  * check_data_file_clashes: Package libgl-mesa-dev wants to install file /srv/home/nitin/build-test-bsps/build-emenlow/tmp/work/emenlow-poky-linux/core-image-sato-sdk/1.0-r0/rootfs/usr/lib/pkgconfig/gl.pc
> >> | 	But that file is already provided by package  * emgd-driver-dev
> >> |  * opkg_install_cmd: Cannot install package libva-glx-dev.
> >> |  * check_data_file_clashes: Package libgl-mesa-dev wants to install file /srv/home/nitin/build-test-bsps/build-emenlow/tmp/work/emenlow-poky-linux/core-image-sato-sdk/1.0-r0/rootfs/usr/lib/pkgconfig/gl.pc
> >> | 	But that file is already provided by package  * emgd-driver-dev
> >> |  * opkg_install_cmd: Cannot install package libgl-mesa-dev.
> >> | ERROR: Function failed: do_rootfs (see /srv/home/nitin/build-test-bsps/build-emenlow/tmp/work/emenlow-poky-linux/core-image-sato-sdk/1.0-r0/temp/log.do_rootfs.49983 for further information)
> >>
> >>
> >> So I think at this point in the 1.4 release, the least disturbing solution will be to revert the commit adding .pc files to emgd recipe.
> >>
> > 
> > OK, so if this is the solution, we need to do test builds and run-time
> > testing on the potentially affected images and arches?  I guess that
> > will now have to include .ipk now too.
> > 
> 
> Ross made a suggestion to just drop the offending .pc file from the emgd
> recipe, it was for something it didn't even provide iirc...
> 
> I can run some builds over the weekend on rage as well once we have
> agreed on a patch.
> 

Right but Nitin said that there were other problems that only reverting
the patch would fix.

Running builds over the weekend is good, but it seems that everything
would need to be re-run-tested as well.  I can also help test once we
have a patch, but probably only emgd targets.

Tom

> --
> Darren
> 
> > Tom
> > 
> >> Nitin
> >>
> >>>
> >>> --
> >>> Darren Hart
> >>> Intel Open Source Technology Center
> >>> Yocto Project - Technical Lead - Linux Kernel
> > 
> > 
> 





More information about the meta-intel mailing list