openmw.org

Interesting.

We’ll need to dig into this some more, but I had a quick look at the code and it at first glance it does appear we have a bug. There’s handling of the source playing or not, but the code seems to expect the offsets to refer to a buffer that is already present. There may be another reason it thinks it’s okay to make this assumption, but at the moment I can’t see it.

We actually have a minor release in the pipes right now and annoyingly it’s too late to change that, but I’ll get sorting this out into our work queue.

For OpenMW, are folk okay to use the workaround suggested below? (Set the offset after the buffer.) I can probably give a bit of advice if that’s helpful (and if you point me at the right bit of code…).

Thanks for the time and effort you’ve put into investigating and reporting this – now we know about it, we can get it sorted out!

Best wishes,

--Richard

AnyOldName3 wrote: ↑20 Aug 2018, 23:25 really Mesa shouldn't have a huge performance penalty for a change which makes no difference on every other GPU driver that's been tested.

Richard wrote: Thanks for the time and effort you’ve put into investigating and reporting this – now we know about it, we can get it sorted out!

raevol wrote: ↑24 Aug 2018, 09:05 I was going to rip on these guys for not following the spec, but they sound really helpful, which is rare! Good on them!