14:55karolherbst: anholt: ohh, one thing I forgot to ask, did you check with some CTS tests if the nir to tgsi work on nv30 causes any regressions?
15:09anholt: karolherbst: as mentioned in the MR, I was unable to test because nv30 hung in an average of like 10 tests, with or without the MR.
15:09anholt: so, like, touch-testing seemed plausible but nothing remotely systematic
15:14karolherbst: anholt: ahh... okay, maybe I try to do it over the next week, because deqp seemed kind of stable, well I also run with -j1 because I know, that sometimes our recovery code is quite broken
15:14anholt: for me, fail was at the console, run a single test, 1-10 times.
15:15anholt: might be a "get a new board" thing?
15:15karolherbst: could be?
15:16karolherbst: anholt: but I have like 5 nv4x GPUs and one nv35 :D
15:16karolherbst: could be that some work better than others