05:37 <icecream95> I wonder if PAN_MESA_DEBUG=gl3 should be enabled by default. Sure, we don't support all of the features required by GL 3.2 and 3.3, but neither do we support all the features of OpenGL 1.0, and *that's* still exposed

06:25 <HdkR> I'm hitting a wackload of games that require GL 4.3, so anything that moves closer to that is better :)

08:03 <icecream95> Oops. I hooked up code to detect buffer overflows and I've already hit a bug, it seems that there are *two* bugs with writing empty buffers to stop prefetching

08:04 <icecream95> "k = ALIGN_POT(k, 2);" My guess is that this is about cache-line alignment, and not actually required

08:06 <icecream95> "Records with continuations must be aligned" oh?

08:18 <icecream95> dEQP-GLES3.functional.depth_stencil_clear.depth_stencil_masked: Crash

08:18 <icecream95> oops

08:22 <icecream95> There appears to be an incorrect assumption that vs->info.attribute_count >= so->num_elements

08:22 <icecream95> broken by util_blitter_clear

08:23 <icecream95> This bug is only two days old!

08:28 <icecream95> I think so->num_elements is wrong

08:32 <icecream95> But create_vertex_elements_state is undocumented so...

08:38 <icecream95> It's worrying how many bugs I've been finding

09:50 <daniels> icecream95: perhaps you could suggest changes or additions to the test framework which would stop those from creeping in

10:00 <icecream95> daniels: Something like a fractional dEQP run with the debug flag I added for detecting overflows?

10:02 <daniels> yeah, certainly

14:48 <robclark> or fractional run w/ asan