[Mesa-dev] DRI3/Present fixes for Mesa 10.3+ (v2)

[Axel Davy]

On Fri, 5 Dec 2014, Mario Kleiner wrote:

>
> Wrt. 2/5: It's a bit ambiguous how to read that bit of the spec, and i agree 
> that one could read it in a way that the current mesa dri3 behaviour is not 
> (completely) violating the spec. When we implemented the DRI2 version, we 
> understood it in the way i want to restore with 2/5. The first reason is 
> because the DRI2 / patch 2/5 interpretation makes the OML_sync_control 
> extension very useful and robust for swap timing, whereas the alternative 
> reading makes the extension borderline useless / its use somewhat fragile due 
> to the race described in the commit. The second reason is backwards 
> compatibility: It would be awesome not to break timing sensitive apps written 
> against DRI2, especially given that users of those apps will certainly not 
> understand why a simple distribution upgrade or graphics stack update pushed 
> by the distro can suddenly cause such regressions.
>

Another thing about 2/5 is that we don't want that querying the current 
msc perturbs the next target_msc when calling present_pixmap when 
swap_interval is not 0.

let's say we have.
-> msc x
last presentation succeeds
-> msc x+1
we query current msc
we make new presentation, target msc y.

Without patch 2/5, the calculus of y will be done on the basis that last 
presentation was done at x+1 instead of x.

Axel Davy