00:15 HdkR: AI writing unittests? My job is in danger!
04:27 mupuf: alyssa: reviewing tests is usually more self contained, but yeah, no blind trusting!
13:38 karolherbst: glehmann: so I have a couple of range analysis optimizations I want to land and was wondering if you have better ideas on how to achieve that, e.g. that one: https://gitlab.freedesktop.org/karolherbst/mesa/-/commit/a36f6b94ce472819654b7f9f27c47d9c503b5770#line_203b86d7d_A4321
13:39 karolherbst: there is a bit more in the commit, but the rough idea is: 1. set no_unsigned_wrap where possible 2. merge ushr(ishl) where safe
14:40 karolherbst: fyi: I've written code to make tsan properly handle simple_mtx and it seems to work out quite nicely :)
14:52 llyyr: there were some issues with util_queue with tsan too
15:08 karolherbst: yeah.. I'm seeing them
15:09 karolherbst: I think I can fix them the same way. As valgrind, tsan has annotations for locks
15:11 karolherbst: though might be better to suppress that one, because it appears to be a "non atomic init" vs "atomic access" situation
15:11 karolherbst: and also maybe should just use C atomics here
15:11 karolherbst: dunno...
15:22 llyyr: i saw those a few months back but my brain was too small to figure out the issue and dog ate my bug report
15:24 karolherbst: anyway: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40421