Time  Nick           Message
00:30 lhofhansl      (OK... Tried that. It's no longer an improvement, I think.)
08:12 BuckarooBanzai !tell lhofhansl if you need help testing the performance related PR's ping me up please, i like your work :)
08:12 ShadowBot      BuckarooBanzai: O.K.
13:32 padresmurfa    Could anybody help me with setting up the Minetest development environment in Docker on a mac? I believe I have a decent setup for building and running the server, but I'm less certain building an OSX client in isolation from the host. Is anybody doing this already, and if so, willing to give me some tips or pointers to docs/code?
13:36 sfan5          none of the core devs use a mac, so I suggest you give compiling in docker a try and ask if you encounter any difficulties, then someone might have an idea
13:38 padresmurfa    ok, will give it a whirl. Thanks :)
13:45 rubenwardy     might be a good idea to try xcode as well
13:49 padresmurfa    Wouldn't that imply building on my host mac? I can't allow building Minetest to make any modifications to my mac, as that would have the potential to break my work environment.
17:50 zughy[m]       uhm, shouldn't MT go into feature freeze?
17:55 rubenwardy     I'm not able to be active currently
20:17 Krock          will merge #10402 in 15 minutes
20:17 ShadowBot      https://github.com/minetest/minetest/issues/10402 -- Add sound to press events of some formspecs elements by pyrollo
20:24 sfan5          one thing: s.sound != "" should be !s.sound.empty()
20:24 sfan5          i'll leave it up to you whether you consider that change important enough to make it
20:26 Krock          will do
20:31 Krock          pushing
20:58 Krock          someone isn't used to the PR merge workflow until 2018. https://github.com/minetest/minetest/pull/10402#issuecomment-732420868
21:40 lhofhansl      Another request for #10637. :)
21:40 ShadowBot      https://github.com/minetest/minetest/issues/10637 -- Avoid generating the same chunk more than once with multiple emerge threads by lhofhansl
21:41 lhofhansl      Without it multiple emerge threads are not that productive since they do unnecessary duplicate work and Lua on_generated is called multiple time for the same mapchunk.
21:43 sfan5          lhofhansl: regarding slow map saving: is compression still a bottleneck at exit when you fix the suspected setTimestamp bug?
21:44 lhofhansl      Yes it is.
21:44 lhofhansl      There are two save cases:
21:45 lhofhansl      1. Saving blocks marked with MOD_STATE_WRITE_NEEDED periodically
21:46 lhofhansl      2. Map::timerUpdate. Where blocks are unloaded and saved with marked at least with MOD_STATE_WRITE_AT_UNLOAD
21:46 sfan5          ah so the timer is likely updated at runtime anyway
21:47 sfan5          which means fixing the setTimestamp bug would only affect few blocks
21:47 lhofhansl      Well, it would avoid automatically every single loaded block as MOD_STATE_WRITE_AT_UNLOAD.
21:48 lhofhansl      (Reminds me to file a PR for that)
21:48 sfan5          true, but don't most of blocks ever loaded get their timestamp updated via the usual ways?
21:49 lhofhansl      If they are active, yes. Every 60s in that case. But that's a subset of the blocks only.
21:49 lhofhansl      Would be nice to have SQL schema that stores the timestamp separately so that we do not have re-serialize the entire block just to update the timestamp.
21:52 lhofhansl      10637 is kinda an important fix. Without it, I'd say we should disallow multiple emerge threads.
21:53 sfan5          the default is still one
21:55 lhofhansl      Yep. And the default is pretty slow. :)
21:56 lhofhansl      Do you have concerns with that one?
21:57 sfan5          no concerns, just needs some careful testing
22:02 lhofhansl      I think it does not change the behavior. In the single threaded case when multiple blocks are generate one of the them "wins", it will cause the chunk to be generated. Other blocks to be generated in that chunk are then loaded from disk instead.
22:03 lhofhansl      The patch makes the behavior the same for the multithreaded case. That's it actually.