[OE-core] [PATCH v3 2/3] mesa: Replace dri-native with dri for native and nativesdk classes

Alexander Kanavin alex.kanavin at gmail.com
Wed Feb 20 17:50:26 UTC 2019


On Wed, 20 Feb 2019 at 18:35, Otavio Salvador
<otavio.salvador at ossystems.com.br> wrote:
> > On Wed, 2019-02-20 at 15:47 +0100, Alexander Kanavin wrote:
> > > On Wed, 20 Feb 2019 at 13:16, Fabio Berton
> > > <fabio.berton at ossystems.com.br> wrote:
> > > > -DRIDRIVERS = "swrast"
> > > > -DRIDRIVERS_append_x86_class-target =
> > > > ",radeon,r200,nouveau,i965,i915"
> > > > -DRIDRIVERS_append_x86-64_class-target =
> > > > ",radeon,r200,nouveau,i965,i915"
> > > > +DRIDRIVERS_class-native = "swrast"
> > > > +DRIDRIVERS_class-nativesdk = "swrast"
> > > > +DRIDRIVERS_append_x86_class-target =
> > > > ",r100,r200,nouveau,i965,i915"
> > > > +DRIDRIVERS_append_x86-64_class-target =
> > > > ",r100,r200,nouveau,i965,i915"
> > >
> > > This produces:
> > > > configure: error: classic DRI driver 'r100' does not exist
> > >
> > > radeon has been renamed to r100 in meson builds, until then it should
> > > be still 'radeon'.
> >
> > Its getting a bit frustrating/tedious to kill all the builds and drop
> > the mesa patches again :(
>
> We are doing our best. It is unfortunate it ended being a
> "one-big-do-everything-together" patch. This should have been done in
> steps since the beginning.
>
> Fabio prepared the new version. We hope it finishes this patchset.

Before you submit any new version, here's another issue, not sure
what's causing it:
https://autobuilder.yoctoproject.org/typhoon/#/builders/65/builds/312/steps/7/logs/step1b
(line 14055 onwards).

'Steps' approach works when each step can be tested quickly. Running
tests on the autobuilder is an expensive, time-consuming operation
(both for the machines, and for RP dealing with the outcome), so those
rounds should be minimized. It is for this reason I did recent perl
and python major version upgrades and recipe rewrites both as single
mega-patches, and not as series - this ensured the minimum possible
amount of testing rounds. Fortunately I also didn't have to convince
anyone about it.

Alex


More information about the Openembedded-core mailing list