<p dir="ltr"><br>
Le 25 oct. 2015 00:00, "Trevor Woerner" <<a href="mailto:twoerner@gmail.com">twoerner@gmail.com</a>> a écrit :<br>
><br>
> On 10/20/15 05:37, Koen Kooi wrote:<br>
> >> Op 20 okt. 2015, om 11:32 heeft Nicolas Dechesne <<a href="mailto:nicolas.dechesne@linaro.org">nicolas.dechesne@linaro.org</a>> het volgende geschreven:<br>
> >><br>
> >><br>
> >> The issue is that mesa is arch specific, not machine specific. So if<br>
> >> you start mixing machines from the same arch, we are getting<br>
> >> unexpected/wrong rebuilds.<br>
> > In this specific case I think we should enable freedreno for armv7/armv8 in OE-core, but the problem remains for other recipes like this.<br>
><br>
> Are you suggesting that every armv7 device should have their mesa tuned<br>
> for freedreno and freedreno added to their filesystem? That seems quite<br>
> overly-broad!</p>
<p dir="ltr">No. Only building the driver. It won't be installed unless you add it into your image.</p>
<p dir="ltr">><br>
> On the surface[1] it would seem as though the decisions as to how one<br>
> builds mesa and which graphics solution to install is more similar to<br>
> the decision as to which DEFAULTTUNE to use, rather than which<br>
> architecture is involved?<br>
></p>
<p dir="ltr">DEFAULTTUNE is a generic config that impacts all packages. What Koen proposed is to change mesa to build more drivers by default.</p>
<p dir="ltr">><br>
><br>
><br>
><br>
><br>
><br>
> [1] which is my way of saying "I don't have the experience of actually<br>
> having done this, unlike you people" ;-)<br>
</p>