Time Nick Message 08:57 sfan5 does anyone feel like looking at #12536 08:57 ShadowBot https://github.com/minetest/minetest/issues/12536 -- Misplaced leaves with visual_scale > 1 08:58 sfan5 also do we want #12515 for 5.6? 08:58 ShadowBot https://github.com/minetest/minetest/issues/12515 -- Add zstd minetest.compress support by 20kdc 12:41 erle sfan5 any idea what minetest.compress is being used for currently except cobbling together PNGs and compressing item meta? 12:44 sfan5 how would I know 12:45 sfan5 I don't know of any uses in mods I come into contact with at least 12:45 erle i thought you looked at that PR for some reason 12:45 erle rubenwardy could i have another CDB dump so i can investigate? or some way to programmatically search CDB for minetest.compress? 12:51 rubenwardy https://content.minetest.net/zipgrep/3c39b4a6-acc6-4bcb-82af-5bb3bc64bef9/ 12:51 erle thanks 12:52 MTDiscord Minetest.compress is also used for stuff like map storage to files by some mods 12:53 rubenwardy I don't think it's worth prioritising that PR for 5.6. But if someone else wants to review it before the deadline they can 12:53 rubenwardy I'd rather have #7865 12:53 ShadowBot https://github.com/minetest/minetest/issues/7865 -- Add crosshair support for Android by srifqi 12:53 erle rubenwardy, are you using grep to find the lines? if so, “git grep -W” can show the function context of the search result (not saying i want it, but maybe that knowledge is helpful if you ever want to improve the results page. 12:53 erle ) 12:53 rubenwardy I'm using `zipgrep` 12:54 rubenwardy https://linux.die.net/man/1/zipgrep 12:54 rubenwardy it's passed to egrep which uses grep -E so should support similar args 12:55 erle ah i used that as well 12:55 erle i have seen the function context thing so far only in git grep 12:55 erle other greps don't have it afaik 13:07 rubenwardy How do I open a new modalmenu from guiFormspecMenu? 13:07 rubenwardy I suppose I'd need to pass in a callback/event handler, which then changes the menu 13:16 Zughy[m] I'd rather have the dual weilding, considering how much the timing is fortuitous for Minetest thanks to Microsoft. You won't ever have an occasion like this again, so please put as many features as you can that can close the gap with MC to pass a message of caring. People won't wait years, only devs will. And you need players, not more C++ geeks 13:21 erle i am not entirely sure what part of the zstd marketing blurb is a lie, but i just tried to figure out the optimal compression level for zstd in terms of minetest content and i was not able to outperform zlib with the default settings for both in terms of compression for a) text below 300 bytes b) ~30kb binary data c) a ~94kb html page. the zstd result is consistently about 5% larger. for text, zstd handily beats zlib when you 13:21 erle specify maximum compression, but then it takes 25 times as long (the difference was 0.01s vs >0.25s to compress the html page). nevertheless, i think the default compression level of zstd should be adjusted to somewhere where it actually beats zlib for content that people are compressing. 13:22 erle if anyone of you has tried to figure out the optimal compression level already, please share 13:23 sfan5 I'm glad erlehmann is here to start discussion about irrelevant things again 13:24 erle uh, i just want to make sure that users do not actually get something worse 13:24 erle i think there *is* a compression level where it beats zlib, it is just not the default one 13:24 sfan5 how users use the API and which tradeoffs they want to make is up to them 13:24 sfan5 we shouldn't mess with library defaults 13:31 erle i think it is relevant if it turns out we need more knobs than “compression level” 13:38 erle okay, according to this https://github.com/facebook/zstd/issues/1134 the problem is known and it depends heavily on what you want to compress. in the example, nothing below 4kb compressed better with zstd than with zlib. 13:40 erle and it makes sense, as they do not care about small files. what “small” is is what i am trying to determine, since the meta i want to compress will not be much more than a few tens of kilobites maximum. 13:40 MTDiscord It would be nice to sneak in any prs that don't have conflicts and are approved (dual wielding? Texture mods?) before the freeze if possible, like zughy said this and next release are opportune times to take advantage of the user influx 13:41 MTDiscord My 2 cents, carry on 13:41 erle sfan5 regarding library defaults: zstd aims to be as fast as zlib with default compression settings. so by default, you seem to gain not much, except the 5% penalty (i am not compressing stuff >100kb in any mod, so i did not test that). 13:41 rubenwardy in other news, #12367 13:41 ShadowBot https://github.com/minetest/minetest/issues/12367 -- Add minetest.get_player_window_information() by rubenwardy 13:42 MTDiscord That would certainly be nice to get in 13:42 MTDiscord But I wouldn't be sad if it had to wait for next release 13:43 erle you have a checkmark at “resist fingerprinting”, how did you solve it? 13:44 rubenwardy if development is particularly active, we could always do releases sooner. Like after 3-4 months rather than 6 13:44 rubenwardy we're not 13:44 rubenwardy you can already use IP addresses 13:44 rubenwardy fingerprinting also requires a lot of different metrics to be reliable, which we don't have 13:44 sfan5 but you're just now adding another one 13:44 erle yes 13:45 MTDiscord Thems the breaks 13:45 erle every bit is useful 13:45 erle anyways, someone™ will have fun making servers handle weird numbers 13:46 erle rubenwardy i understand “i do not want to solve it” but then the checkmark should not be before “resist fingerprinting” 13:46 erle especially since we live in a world in which evil actors can and will gobble up any info and use it against anyone 13:47 erle (NSA using google tracking cookies etc.) 13:48 MTDiscord Disregarding a faulty checkmark, the pr is ready 13:52 erle btw, i figured out how people use zstd for small data (smaller than ~32kb) – they provide dictionaries, because that data is often uniform. if you do that, it indeed outperforms zlib by A LOT. if you do not do that, it tries to “learn” the structure of the data and the tradeoffs can be suboptimal for quite a long time. 13:53 erle (i.e. if i would use minetest.compress in a mod using zstd, i would want to provide a dictionary, because i am only compressing stuff <100kb anyways)