Minetest logo

IRC log for #minetest-hub, 2019-03-03

| Channels | #minetest-hub index | Today | | Google Search | Plaintext

All times shown according to UTC.

Time Nick Message
00:05 paramat joined #minetest-hub
00:13 kaeza joined #minetest-hub
00:37 Sokomine joined #minetest-hub
00:51 sofar IhrFussel: I've been actually skeptical of the claim that it isn't threadsafe, but I have no data or evidence to support that MT's use of it is thread safe (it could be an MT implementation issue after all)
00:52 paramat i'll check with nerz, see why he wrote that
00:53 sofar well, here's the catch
00:53 sofar setting the sqlite threading mode is done at *compile time* of sqlite
00:54 sofar in most distro's that means you're using the distribution libsqlite3.so
00:54 sofar so you might be safe on one distribution, but not another
00:54 sofar I just checked clearlinux and we don't change the default, so, on clearlinux, sqlite3 should be serialized and thus threadsafe for minetest
00:55 sofar but, you have to check (every | your own) distro, or compile sqlite3 statically
00:59 sofar heck you can ask during runtime if sqlite3 was built thread safe with sqlite3_threadsafe()
01:15 benrob0329 joined #minetest-hub
01:17 paramat ah hm
01:18 paramat so MT code should check it
01:33 kaeza joined #minetest-hub
01:52 Miner_48er joined #minetest-hub
03:11 paramat joined #minetest-hub
04:34 diemartin joined #minetest-hub
06:52 GreenDimond joined #minetest-hub
07:44 Emerald2 joined #minetest-hub
08:26 CWz joined #minetest-hub
09:14 IhrFussel sofar, yeah now I realize it depends on the distribution's binary of sqlite ... but not sure why one would choose another mode by default, makes little sense to me
10:21 _Xenon joined #minetest-hub
10:50 Fixer joined #minetest-hub
12:33 calcul0n joined #minetest-hub
12:49 Unarelith joined #minetest-hub
13:06 Sokomine i still wish mapblocks would store when they where created, when last changed, how many blocks digged and placed by players. that could fascilitate making backups and removing mapchunks exploring players visited once and which just take away disk space
13:10 calcul0n joined #minetest-hub
13:41 Fixer Sokomine: even easier way is don't save mapblock if no changes were made to it
13:49 Sokomine fixer: hmm. what about mapblocks close to spawn that are never changed? compute them anew all the time? that's no real solution either...
13:50 Fixer Sokomine: no, it is different, once you change it once, it get saved all the time
13:51 Sokomine yes. still, even in densely populated areas, there may be many mapblocks that didn't receive a single change. the surface of the world is big
13:52 Sokomine there's another problem: mapgen works in mapchunks. so if you don't save part of it, it'll all generate anew. which isn't exactly desired if other mapblocks in the chunk contain changed data
13:53 Sokomine and if your mapgen changes inbetween, you're bound for even more trouble. what you suggest may be a solution for cases where disk space is extremly rare
13:54 Sokomine if there where a date of creation and last change as a reference, map shrink tools might identify hotspots and keep mapblocks around those hotspots intact, even taking care of how mapgen as such works
13:56 Sokomine might also be intresting to keep a history of how a server developed. if you have a date as a reference and a database, you can store more than one mapblock with the same coordinates and switch back to previous incarnations if necessary
13:58 Sokomine just...players digging tunnels in mines and placing the occasional torch, cobble, dirt or whatever to climb up or down is not particulary intresting
14:25 aerozoic joined #minetest-hub
14:56 entuland joined #minetest-hub
15:47 _Xenon joined #minetest-hub
17:02 Krock joined #minetest-hub
17:08 Krock IhrFussel: u_i is running stable, I guess? I'd like to merge the PR tomorrow; or it would have to wait until next weekend
17:27 benrob0329 joined #minetest-hub
17:39 IhrFussel Krock, it didn't cause crashes at least since I updated it and no player complained about something broken
17:46 Krock nice, thank you :)
18:12 rdococ joined #minetest-hub
18:17 IhrFussel Krock, it can still cause lags of up to about 1 second...but those are probably just exceptions
18:18 IhrFussel - on_player_receive_fields[2] .......................  |         1 |   1118034 |       437
18:19 Krock check out the average value. is it the 437us?
18:21 _Xenon joined #minetest-hub
18:23 _Xenon joined #minetest-hub
18:25 IhrFussel Yes but that one also counts when the formspec is just ""
18:26 IhrFussel So it doesn't only run when the fields match the function
18:26 IhrFussel I will restart in 15 minutes...will add a custom timer now
18:31 IhrFussel Krock, which var can I use to get the item crafted? I wanna log it as 'PLAYER crafted '..amount..' '..item..' within...'
18:33 IhrFussel I mean the itemstring like 'default:bookshelf'
18:35 IhrFussel Maybe 'output'?
18:35 IhrFussel local output = unified_inventory.current_item[player_name] or ""
18:35 Krock craft.output
18:36 Krock or output, yes.
18:36 Krock two times the same thing
18:39 _Xenon joined #minetest-hub
18:40 IhrFussel Resrarting now...then we have actual numbers for the funtion running
18:40 IhrFussel Restarting*
18:40 IhrFussel function* ... keyboard dying
18:44 IhrFussel When I take away a few stacks from the craft grid and press the buttons it takes much longer
18:45 IhrFussel 2019-03-03 19:42:46: ACTION[Server]: CRAFT TIMER: IhrFussel crafted 30 default:bookshelf within 3524 µs
18:45 IhrFussel This is when all my items are in my inventory and I press buttons
18:46 IhrFussel If I now shift click a few stacks from the craft grid back into my inventory and press the button again
18:46 IhrFussel 2019-03-03 19:44:07: ACTION[Server]: CRAFT TIMER: IhrFussel crafted 30 default:bookshelf within 360646 µs
18:48 IhrFussel 'crafted' is maybe wrong...should be 'moved'
18:50 IhrFussel Yep... as soon as I take one stack out of the craft grid again and press button the execution time suddenly jumps
18:52 IhrFussel Will make a video of it
18:55 Calinou how do I effectively load intllib's S, NS into many files in the most DRY way possible?
18:55 Calinou https://github.com/minetest-mods/intllib/blob/master/doc/developer.md doesn't explain it, I feel like it should be clarified
18:56 Krock local S, NS = dofile(MP.."/intllib.lua");  mymodnamespace.S, mymodnamespace.NS = S, NS;  then re-use in the other files
18:56 Calinou thanks :) that's what I thought
18:57 Calinou I had tried this:
18:57 Calinou local S, NS = dofile(modpath .. "/intllib.lua")
18:57 Calinou moreblocks.intllib = {S, NS}
18:57 Calinou but that doesn't work as they're considered table values
18:57 Krock you might want to have a fallback, though
18:57 Krock well it does, but then you'll use
18:57 Krock local S = moreblocks.intllib[1]  on top of each file
18:58 IhrFussel Krock, https://ihrfussels-server.de/craftdemo.mp4
18:59 Calinou right now, I need this at the top of each file:
18:59 Calinou local S = moreblocks.S
18:59 Calinou local NS = moreblocks.NS
18:59 Calinou so I don't have to use moreblocks.S() all the time, which is quite verbose
18:59 Calinou (I can remove NS if a file doesn't need it)
19:00 Krock IhrFussel: 2ms sound about right
19:00 Krock oh shit
19:00 Krock why the heck is that happening. Every time?
19:01 IhrFussel Yes as soon as the code needs to skip over empty craft grid slots
19:01 Krock did you mean: skip over occupied slots?
19:02 IhrFussel The 2nd move action in the video shows that I took 1 slot away from the grid which resulted in the high time
19:02 IhrFussel I guess the more slots I take away the higher the time...let me check
19:03 IhrFussel YES
19:03 IhrFussel I reached 416ms with only 1 slot of books left before clicking the 'to craft grid' button
19:04 Krock https://github.com/minetest-mods/unified_inventory/pull/123/commits/2a729185
19:04 Krock should be faster with that, as it skips the "groups" check
19:05 Krock moment
19:05 Krock amending the commit for further improvement
19:05 Krock nvm, I thought that I didn't skip empty slots. false alert
19:05 IhrFussel You really think that's the cause? Cause it clearly only happens when I take away stacks from the craft grid so that there are 'holes' before clicking the button
19:06 Krock it's weird indeed
19:06 IhrFussel I will apply your lines and check again
19:07 Calinou alright, updated intllib support in my mods, thanks :)
19:07 Krock will be off in ~10 minx
19:07 Krock np
19:07 Calinou I'll switch to .po files when I have time
19:07 Calinou (and use string formatting more liberally)
19:11 IhrFussel Nope, still same... as soon as I take away 1 slot from the craft grid the time jumps to 100+ ms
19:12 IhrFussel Even when I only press '10' it's still 90ms
19:12 Krock weird
19:13 IhrFussel No engine bug I hope
19:14 IhrFussel For some reason the mod takes a lot longer to check SOME empty stack slots and fill them vs all slots being empty
19:15 Krock the amount of time spent must be proportional to the empty slots
19:16 IhrFussel WAIT...it doesn't happen when either the first or last slot are non-empty
19:16 Krock ????
19:16 IhrFussel Making another quick video one sec
19:19 Krock that's not logical at all
19:19 IhrFussel https://ihrfussels-server.de/craftdemo2.mp4
19:19 IhrFussel It's weird...but it seems to depend on the slots I remove
19:21 IhrFussel The last one was by another player, my time was 3 ms
19:22 IhrFussel Last line in terminal*
19:23 Krock will debug further next weekend
19:23 Krock could be some nasty C++ bug or table reference
19:39 kaeza joined #minetest-hub
19:51 GreenDimond joined #minetest-hub
20:22 garywhite joined #minetest-hub
20:23 garywhite joined #minetest-hub
20:25 Unarelith joined #minetest-hub
20:58 FrostRanger joined #minetest-hub
21:00 kaeza Greetings.
21:15 Fixer kaeza: Hi
21:33 T4im joined #minetest-hub
21:37 pauloue joined #minetest-hub
21:39 benrob0329 Hello
21:58 T4im joined #minetest-hub
22:23 _Xenon joined #minetest-hub

| Channels | #minetest-hub index | Today | | Google Search | Plaintext