Bug 336854 – x11-libs/libva-1.0.7 (new ebuild)

• Status: RESOLVED FIXED
• Product: Gentoo Linux

[Description Sergey Ilinykh 2010-09-11 19:31:52 UTC]

i don't know how stable it is but some distros already have it.

Reproducible: Always

[Comment 1 MT 2010-10-21 16:17:59 UTC]

Created attachment 251457 [details]
x11-libs/libva-1.0.5.ebuild

An updated and improved ebuild for x11-libs/libva

[Comment 2 Michał Górny 2010-10-21 16:20:33 UTC]

Thanks for the ebuild. Could you attach a diff against the current ebuild as well?

[Comment 3 MT 2010-10-21 16:39:06 UTC]

Created attachment 251461 [details, diff]
libva-1.0.5.ebuild.patch

Sure Michal, here the ebuild patch against the current ebuild (libva-0.31.1_p4.ebuild).

p.s.: I have done a fix to the new ebuild in the meantime, there was a patch (files/libva-dont-install-tests.patch) used by libva-0.31.0*.ebuild but not by libva-0.31.1*.ebuild (the ebuild that I have used as a skel for the new one) and now that patch is applied too as it avoids the installation of the tests (completely useless for end users).

[Comment 4 MT 2010-11-06 10:33:30 UTC]

The submitted ebuild works as it also for v1.0.6, it's enough to rename as libva-1.0.6.ebuild.

[Comment 5 emil karlson 2010-11-06 15:35:59 UTC]

Cgit seems to recreate the tarball for each fetch with different gzip timestamp so manifest is always invalid, so the SRC_URI might be a problem.

For some reason vainfo reports library version 0.31.1.

Otherwise no regressions for i965 driver for libva-1.0.6 on GMA4500MHD (mpeg2 only).

[Comment 6 Sebastian Mueller 2010-11-16 22:51:04 UTC]

I used that ebuild for libva-1.0.5 to get all version as shown in 
http://intellinuxgraphics.org/2010Q3.html and have no problems so far. 

Thanks for the ebuild.

Is the problem with the tarball still exists?

[Comment 7 MT 2010-11-17 08:13:49 UTC]

(In reply to comment #5)
> Cgit seems to recreate the tarball for each fetch with different gzip timestamp
> so manifest is always invalid, so the SRC_URI might be a problem.

(In reply to comment #6)
> Is the problem with the tarball still exists?

For now never mind the Manifest problem, it can be easily fixed by pushing a copy of the tarball into the Gentoo's main distfiles and changing the ebuild SRC_URI to directly grab the Gentoo copy instead of the one provided by upstream (as done, for the same reasons, by a lot of other ebuilds into the portage tree). This fix will be made by the Gentoo devels when the ebuild will be submitted in the tree.



(In reply to comment #5)
> For some reason vainfo reports library version 0.31.1.

This is an upstream bug, can you report it to the upstream bugtracker/mailing-list? thank you.

[Comment 8 Alexandre Rostovtsev (RETIRED) 2011-01-10 03:40:55 UTC]

Created attachment 259434 [details, diff]
libva-1.0.7.ebuild.patch (against current libva-0.31.1_p4.ebuild)

libva-1.0.7 is out, and Intel considers it stable enough to be part of their 2010Q4 graphics package release (see http://intellinuxgraphics.org/2010Q4.html).

Changes relative to Toffanin's 1.0.5 ebuild:
* updated libdrm dependency;
* tar.bz2 instead of tar.gz for smaller downloads;
* do not delete .la files if USE=static-libs.

[Comment 9 Alexandre Rostovtsev (RETIRED) 2011-01-10 03:42:12 UTC]

Also, since libva-dont-install-tests.patch was deleted from the main portage tree, you will need to grab it from http://sources.gentoo.org/cgi-bin/viewvc.cgi/gentoo-x86/x11-libs/libva/files/libva-dont-install-tests.patch?revision=1.2

[Comment 10 MT 2011-01-12 11:45:13 UTC]

(In reply to comment #8)
> libva-1.0.7 is out, and Intel considers it stable enough to be part of their
> 2010Q4 graphics package release (see
> http://intellinuxgraphics.org/2010Q4.html).

Thank you for the new ebuild, it's working well here.

[Comment 11 Luke-Jr 2011-01-21 04:08:27 UTC]

FWIW: libva 1.0.7 is required for Sandy Bridge video decoding accel

[Comment 12 Tomáš Chvátal (RETIRED) 2011-03-08 12:28:30 UTC]

1.0.10 added to main tree.