Time Nick Message 05:28 sfan5 idea for /clearobject taking too long, eating all ram fix: /clearobjects only clears the loaded chunks, the unloaded ones will get an "clearobejcts=1" flag and the objects will be cleared once they are loaded 05:29 VanessaE sounds fine to me 05:29 VanessaE if anyone cares to actually code it 05:30 kahrl it could even be done without loading all mapblocks to set a flag on them 05:31 sfan5 yeah 05:31 kahrl store a setting in the map metadata that stores when the clearobjects command was run the last time 05:31 sfan5 BUT wouldn't that break the map format? 05:31 kahrl map metadata is extensible 05:31 sfan5 oh right, good 05:32 kahrl and mapblocks already store the time they were saved 05:38 kahrl hmm, map metadata is currently only used for mapgen stuff so it's tied into the EmergeManager 05:39 kahrl but I guess env_meta would work 05:39 VanessaE the problem isn't how long it takes, the problem is the fact that it doesn't take 10GB of RAM to store one mapblock (i.e. it should never have to keep more than one in memory, plus whatever's already loaded) 05:42 kahrl the intent between /clearobjects may also have been to immediately reduce the size of the map on disk 05:42 kahrl that's why it would immediately load and modify all blocks 05:44 VanessaE perhaps 05:49 celeron55 the intent of /clearobjects is to immediately get rid of all objects (many objects in random places can arise from bugs) 05:49 celeron55 it should be just fixed so that it doesn't load everything into memory at the same time 05:52 kahrl suggestion for an improved /clearobjects: http://paste.dy.fi/Ac3 05:55 kahrl /clearobjects -i would work as it does currently but ideally fixed as celeron55 said 06:03 kahrl I haven't done stuff in map.cpp in a long time but I might try to add the unloading code for /clearobjects -i right now 06:20 hmmmm hey guys 06:20 hmmmm i think it's great somebody's doing something about clearobjects, 06:20 hmmmm http://dev.minetest.net/TODO#Other 06:20 hmmmm it was kind of on the todo list for quite a while 06:21 hmmmm and yeah, env_meta is the best place to put that IMO 06:22 VanessaE evening, hmmmm. 06:25 celeron55 i don't agree at all with the at-load clearing of mapblocks 06:25 celeron55 it's useless 06:25 hmmmm useless?? 06:25 celeron55 or what would people use it for? 06:25 hmmmm it'd fix clearblocks 06:25 celeron55 clearblocks? 06:26 hmmmm clearobjects 06:27 celeron55 not at least for it's original usage, not at all 06:28 celeron55 if i would have had that kind of /clearobjects on the server flooded with DMs, it would have been completely useless 06:35 kahrl that's why I won't just implement directly what sfan5 suggested 06:35 kahrl sometimes clearing at-load is useless so I want to give users the option 06:41 hmmmm oh, you're talking about that idea 06:41 hmmmm i meant putting a limit on the number of mapblocks that can be loaded at once, and if it hits the limit, in order to load that new one it'd have to clear the oldest one or something 06:54 kahrl I'm not sure that would work well with clearobjects 06:54 kahrl All the mapblocks that are newly loaded by clearobjects would be newer than the ones that were already loaded 06:55 kahrl So after clearobjects all the "important" mapblocks (near players) would be unloaded 06:55 hmmmm yeah, it would wipe out what was loaded already, you can easily make an exception for this if you'd like 06:56 hmmmm you'd just have to add a new flag or something 06:56 kahrl I think even just calling addRef/dropRef on the already loaded blocks at the beginning and end of clearobjects would do the trick 06:57 hmmmm what are those? 06:57 kahrl each mapblock has a reference count 06:57 kahrl I think it's currently used for the draw list 06:57 hmmmm they'd have to be unloaded immediately though 06:58 kahrl yeah obviously just doing that wouldn't unload the rest of the mapblocks 06:58 kahrl s/addRef/refGrab, s/dropRef/refDrop 06:59 celeron55 hmm, so if there was a "clear objects in older blocks than " in env_meta.txt, it could simulate the result of a full /clearobjects 07:48 kahrl for the glob idea I just wrote a portable glob http://paste.dy.fi/8nB 07:49 kahrl needs some testing but seems to work 07:58 celeron55 or you could just directly pass it to lua's pattern matcher 8) 07:58 kahrl hehe 07:58 kahrl well programming this was fun anyway ;) 07:58 celeron55 it supports a good portion of what regexes usually do 12:07 PilzAdam celeron55, if someone throws a massive amount of items out of his inventory then he would like to run a /clearobjects for the active area 12:11 Calinou or add a /clearinventory 12:11 Calinou better 12:11 Jordach isnt there a mod for that 12:12 PilzAdam and what happens if I call add_entity() in an unloaded mapblock after its marked to be cleared? will it be removed too when its activated? 12:12 Jordach or even, have a clear all slot which clears all of the item in that slot and in the inv 12:12 Jordach so i could put default:cobble and it clears all the cobble 12:29 sfan5 PilzAdam: add_entity doesn't load an unloaded mapblock, so there will be no entity added 12:35 thexyz or /pulverize 12:35 thexyz wait, i'm a bit slow 12:37 sfan5 "a bit" 12:56 sfan5 has anyone seen erlehmann this/last year 12:58 PilzAdam yes 13:32 rubenwardy !seen erlehmann 13:33 sfan5 rubenwardy: minetestbot is not in here 13:33 sfan5 <MinetestBot> sfan5: Sorry, I haven't seen erlehmann around. 13:33 rubenwardy hmm 13:50 Jordach hes logged in 17:41 sapier how would incremental clearobjects behave if a new object moves from a marked block? 17:41 sfan5 I guess it wouldn't clear the object then 17:42 sapier but the block is still marked as toclear? 17:42 sfan5 But while /clearobjects is running there should be no env steps 17:42 sapier with incremental there would 17:43 sfan5 the MapBlock would remain flagges as toclear 17:43 sfan5 *flagged 17:44 sapier thus the new object will be removed once this block is loaded? 17:47 sfan5 if it moves to a block that was already cleared, no 17:48 sapier that's not what I was asking ;-) 17:48 sfan5 oh.. 17:48 sfan5 it won't be cleared then 17:49 sfan5 (if it moved to a block that is not marked) 17:49 sfan5 *moves 17:49 sapier I don't think so .. I guess it will be cleared although it was created after clearobjects has been called 17:49 sfan5 the toclear flag is intended for unloaded blocks anyway 17:50 sapier yes that's exactly what I'm talking about 17:50 sapier a mob for example can walk from an loaded to an unloaded block beeing unloaded at that moment 17:50 sapier if that block was marked toclear the mob will be deleted once the block is loaded 17:51 sfan5 yeah 17:51 sapier unless the block is loaded to store the mob ... not quite best thing but I'd not be surprised if minetest did handle it that way 18:25 sapier https://github.com/minetest/minetest/pull/724 squashed again ... hope it's good to merge now 18:46 sweetbomber hello 18:47 sweetbomber i have discovered minetest project a few days ago, and i am willing to collaborate 18:49 sapier http://dev.minetest.net/TODO 18:50 PilzAdam https://github.com/minetest/minetest/issues?labels=bug&page=1&state=open 18:51 sweetbomber how can i actively collaborate? I have decent programming skills but i can also improve (G)UI and textures 18:51 PilzAdam code a new features/fix something and then open a pull request 18:51 PilzAdam put I recommend talking about what you want to do first here 18:51 sapier do you know forum already? have a look whats going on there 18:52 sapier https://github.com/minetest/minetest/issues/346 I guess this one is to be closed atm isn't it? 18:53 sweetbomber ill take a look in the forum. the pull requests are made on git or somewhere else? 18:53 PilzAdam on github 21:03 troller after server crash i have env_meta.txt with 0 size 21:03 troller and Automatically selecting world at [/home/proller/.minetest/worlds/world] 21:03 troller terminate called after throwing an instance of 'SerializationError' 21:03 troller Abort trap: 6 (core dumped) 21:05 celeron55 known bug (happens randomly at crashes); just insert something valid in there like this: http://paste.dy.fi/i8U/plain 21:06 troller yes, but something else stops... 21:07 troller wooow 21:07 troller world/players 21:07 troller all files 0 size 21:08 troller 0 bytes in 1736 files 21:08 troller its because they saves every minute or faster 21:08 celeron55 ouch 8) 21:09 troller started after deleting 21:10 celeron55 i'm surprised nobody has fixed the file saving things 21:10 troller nobody knows 8) 21:10 celeron55 people do all kinds of useless stuff but never touch those 21:10 celeron55 all that would need to be done is saving to a temporary file and then moving them with filesystem commands in place 21:10 troller i now no time to it 21:11 troller have no 21:11 sapier lol :-) I didn't even know about that bug 21:11 troller simpler save only on change 21:12 celeron55 troller: it doesn't fix it, just makes problems more rare 21:12 troller on powerloss can lost only last users 21:12 celeron55 but it'd need to be done too (for players, that is) 21:12 troller move can loss too 21:12 celeron55 when MT crashes, it can't 21:13 sapier solutions not solving a problem entirely aren't worth the work ... at least if there IS a solution solving the issue 21:13 celeron55 if you crash your whole computer, nobody can save you anyway so why even mention it 21:13 celeron55 (except that a proper journaling filesystem *will* save you) 21:14 celeron55 (as long as it's done in the move way) 21:14 troller tunefs: gjournal: (-J) disabled 8(