Created attachment 115622 [details]
Hardware state is not reset correctly when a process makes use of a new GL context, after it has used and then destroyed the previous one.
This manifests in a (still uncommitted, https://www.winehq.org/pipermail/wine-patches/2015-April/138974.html) Wine test which fails on nv50. Actually commenting out stream_test(), which makes use of instanced draws with ARB_instanced_arrays, workarounds the issue.
A single process executes both tests, destroying and then recreating a GL context in between. AFAIU specifically the per-instance enable flags are never reset on the HW and that breaks the latter test.
I'm attaching a patch which "fixes" the issue for me but it's not the correct approach.
Created attachment 115623 [details] [review]
Ilia's attempted fix
I'm attaching Ilia's patch for this issue, from IRC. It doesn't work for me (yet) but this should be more like the correct fix.
BTW, I can try to make a standalone, GL testcase for this bug. It might take a while though.
Created attachment 115624 [details] [review]
The attached works for me. It's the patch from Ilia with an additional state copy in nv50_create.
Ah yeah, that makes sense. Forgot that contexts liked to auto-attach themselves. cur_state seems like a dumb name, I'm going to change it to saved_state or something. Also that state needs to be initialized, so I'm going to move the state init bits from that state init function to the screen init. Will send a proper patch for testing tonight-ish.
Just pushed out the fixes to make it work on nv50 and nvc0. nvc0 had the additional bug that 3d-engine-based blits use a vbo, and were also forgetting to overwrite these attributes, leading to a fail even when having the proper "state".
Thanks for reporting!
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct.