Time Nick Message 10:11 Zughy[m] Hello, I need more power on GH if you want me to actually manage every issue. I can't move them, I can't rename them, I can't edit comments. Yes, I know that the next step means I can also push to master, but I also think you know me well enough to know I won't ever do that 10:22 Zughy[m] also, reminder that in a few months the roadmap should be discussed again, as in October it will have been two years since the first one was created (#10461). To make life easier for everyone, this could happen after 5.7 is out 10:22 ShadowBot https://github.com/minetest/minetest/issues/10461 -- Roadmap Brainstorm 10:23 Zughy[m] (the issue was created in October, then it took almost a year to have the roadmap, June 2021) 10:46 MTDiscord Not that I have any say in this, but I'd be fine with you pushing to master as long as you stick to stuff you can properly review (which likely excludes most C++ stuff). 12:06 * kilbith handles the power to Zughy to ban erleshit 12:21 MTDiscord kilbith: Calm the fuck down 12:21 MTDiscord erle has different priorities than you (and probably most of the core devs), I get it 12:21 MTDiscord do we have a way to tell the commit has at wich irrlitch is supposed to build with the 5.5.1 minetest release? 12:21 MTDiscord *hash 12:22 MTDiscord no need for a ban though, erle is definitely still a net plus 12:23 MTDiscord I'm testing some AppImage builds for me to test RC/Dev versions but I failed to compile tag 5.5.1 with irrlitch master 12:23 rubenwardy the release says it 12:23 rubenwardy use 1.9.0mt5 12:24 MTDiscord got it, thanks! 12:32 MTDiscord #12610 12:32 ShadowBot https://github.com/minetest/minetest/issues/12610 -- Cant pickup cart 12:32 MTDiscord if anyone cares to decrease the number of issues ^ 12:34 rubenwardy sounds like it should be made possible to sneak then long tap 12:34 MTDiscord yes it should and already made an issue :) 12:35 MTDiscord #11404 12:35 ShadowBot https://github.com/minetest/minetest/issues/11404 -- Android: Can't use sneak/aux while digging or punching 12:35 rubenwardy right, helpful to point out the duplicate 12:35 rubenwardy found it 12:36 MTDiscord > right, helpful to point out the duplicate removing carts IS possible in android 12:37 MTDiscord so the other issue is not really a bug 12:37 rubenwardy what the issue says isn't necessarily what the issue is about 12:37 kilbith rubenwardy: do you read my feedbacks on CDB? 12:37 MTDiscord > what the issue says isn't necessarily what the issue is about i thought it was about picking up carts 12:38 MTDiscord anyway, it will be a lot better to tweak the way long tapping works 12:38 rubenwardy the user had an issue picking up carts. The cause is due to that bug. So if the previous issue didn't exist, I would have changed the title of the carts issue to "Can't use sneak/aux while digging or punching" 12:38 MTDiscord alright 12:38 rubenwardy kilbith: I drafted a reply but didn't finish writing 12:48 MTDiscord Would it worth sending a Pull Request to get AppImage build as part of the CI in Github repo? I spent a few time doing it in a way that seems to be reasonable to share upstream instead of keeping an unofficial repo for it. 12:50 MTDiscord I can prepare a PR with the adjusted scripts if needed: https://github.com/ronoaldo/minetestclient 13:04 sfan5 there's an app image build on gitlab ci IIRC 13:04 sfan5 just nobody ever uses them 13:08 MTDiscord oh, well, there is indeed one at Gitlab CI. it is a bit hard to find if you're not familiar with the Jobs UI but it is there indeed - so no need to duplicate that. I was a good learning anyway \o/ 13:09 MTDiscord I do find AppImage an easy to grab-and-go, specially useful for new Linux users. Packaged versions from Distros (Debian, Ubuntu) are outdated, and with the single file download it is an interesting option as a canonical/universal release. 13:10 MTDiscord Universal sure, but canonical ... that'd be the snap, right? ? 13:11 MTDiscord definitely! 13:12 schwarzwald[m] sfan5: Lies, all lies! I use the AppImage lol. 13:14 MTDiscord Yeah, I use the Bitwarden (with auto-updates using appimagelauncher from Debian) and it works like a charm. It is the upstream default Linux packaging too. 13:34 MTDiscord There seems to be an issue with the current Gitlab AppImage - the same I faced when trying to build on my machine. It uses on Ubuntu Bionic to fetch dependencies, and it is almost a decade old. 13:35 MTDiscord The mismatch of the Ubuntu version causes some issues like not having all the .so dependencies or not in the correct versions as used during the build stage, or having too old libs in place and causing it to crach when open: https://paste.debian.net/hidden/ef26cb87/ 13:36 MTDiscord schwarzwald[m]: what .AppImage you use? Or do you use it on Ubuntu itself? 13:38 MTDiscord Ronoaldo: Bionic Beaver is 18.04 and "only" 4 years old, meaning LTS is about to expire? 13:38 MTDiscord Or were you referring to the machine whan you stated "and it is almost a decade old"? 13:38 MTDiscord oh, my bad - I tought it was 14 something and rounded up 2 yesr 13:38 MTDiscord ah I see 13:39 MTDiscord then I'm not sure why the AppImage file don't launch properly on my Debian 11 host 13:39 MTDiscord It complains about Open GL drivers and I tought it was about either opengl .so or perhaps it is mesa on host? 13:40 MTDiscord I also think that there needs to be a directive to fix the missing system-wide folders - after that, my .appimage build worked. 13:40 MTDiscord Specifically the path_mappings directive: https://github.com/ronoaldo/minetestclient/blob/main/AppImageBuilder.yml 13:40 MTDiscord I used Debian to match my host build but it should work just fine on .AppImage 13:45 MTDiscord I'll test on a VM to make sure it is not something with my host OS. 13:54 schwarzwald[m] Ronoaldo: I downloaded the AppImage from the Minetest downloads page I think, and I use it on Debian 11. I haven't updated it in quite a while, so it's possible there's a problem only in the newer version. 13:57 MTDiscord thanks! I failed to find it on the site download. I did found an unofficial build on forums; since the github don't have the sources/log of the build, I went cautious on that one. 14:03 rubenwardy the minetest readme doesn't say how to install xinput 14:03 MTDiscord xinput indeed is missing from the AppImage file as well.. perhaps that is the issue? 14:03 rubenwardy what's the package on ubuntu? libxcb-xinput-dev and xinput? 14:04 MTDiscord on bionic I think it is libxi-dev and libxi6 https://packages.ubuntu.com/bionic/libxi6 14:05 MTDiscord From the source package: https://packages.ubuntu.com/source/bionic/libxi 14:05 rubenwardy aga 14:05 rubenwardy *aha 14:05 rubenwardy that is in the list 14:05 MTDiscord I guess it is missing from here, right? https://gitlab.com/minetest/minetest/-/blob/master/AppImageBuilder.yml 14:06 MTDiscord include: directive do not mentions that one. 14:10 MTDiscord it "just worked" on a Ubuntu 18.04 vm fresh install 14:15 MTDiscord It failed the same way on Ubuntu 22.04 (I also had to manually install lib Fuse - it was not a download and run experience as I expected it to be). 14:20 MTDiscord https://imgur.com/a/pb0Yn5v 14:23 MTDiscord Nice, Github artifacts only work for authenticated users, not for guests ? 15:50 MTDiscord yeah, appimage is still messy (at least to what I could understand of it). build on Debian, don't run on Ubuntu. And vice versa! 15:50 MTDiscord so not as portable as I was thinking 15:53 sfan5 "Github artifacts only work for authenticated users, not for guests ?" yes, that's a reason not to use githubs CI 16:08 MTDiscord so far I managed to have either it working on Debian11+Ubuntu22.04 or on Ubuntu18.04 - not able to build/test any appimage working on all options 16:08 MTDiscord all options = all three 16:10 MTDiscord Regardless of the version combo, the seg fault is consistently preceeded by this 'Irrlicht: OpenGL driver version is not 1.2 or better.' 16:31 MTDiscord oh for gods sake, what noise source is Minetest actually using 16:31 MTDiscord it's not reference Perlin 16:35 schwarzwald[m] Who knows, it might be a custom noise algorithm! 16:44 MTDiscord Whenever I read noise I immediately think about someone screaming at the PC to make the noise and generate the world ? 16:54 Calinou that could be a valid source of entropy 16:54 Calinou (like how generating a GPG key requires you to move the mouse around) 17:03 kilbith we could actually drop Perlin noise and adopt std::mt15537 alongside with std::uniform_{int,random}_distribution 17:03 kilbith bonus: you use the STD; cons: it breaks the maps 17:04 kilbith s/D/L 17:04 kilbith * std::19937 17:06 schwarzwald[m] Is Mersenne Twister really a drop-in replacement for a noise algorithm? Perlin noise smooths out the gradients by interpolating between adjacent points in the noise map IIRC. 17:06 MTDiscord i'm in the depths of mapgen and noticed heinous issuews 17:07 MTDiscord like how 3d noise is heinously inefficient in terms of CPU time 17:07 schwarzwald[m] Jordach is dug too greedily and too deep. 17:07 MTDiscord not too deep 17:07 MTDiscord found out even the C++ implementations are a garbage fire 17:08 MTDiscord i might write an interface for Lua that gives access to reference perlin noise as well as engine 17:08 schwarzwald[m] By reference perlin noise you mean the reference implementation as defined by the spec? 17:11 MTDiscord yes 17:11 MTDiscord what we have has axial issues 17:11 MTDiscord certain mapchunk regions have gridding 17:11 schwarzwald[m] That's awesome - that you're adding a good Perlin noise implementation, not that we have issues. 17:12 MTDiscord it is quite amusing that the more i dug into the C++ mapgen the more cockroaches i found 17:12 schwarzwald[m] If anybody knows why we didn't use a regular Perlin noise in the first place, I want to hear it lol. 17:14 schwarzwald[m] Jordach, if you reach the deepest possible depth in a Minetest world at a specific set of coordinates that vary slightly due to undefined behavior, you find where the mapgen code is sealed up in mossy cobble, and the walls covered with an inch thick layer of giant cockroaches. 17:15 schwarzwald[m] And the game will crash because it can't handle all those entities at once. 17:15 MTDiscord spoiler alert 10s of thousands of lua objects don't bother the server 17:16 MTDiscord the suspiciously high limit doesn't get upset that mucch 17:18 schwarzwald[m] Ah. Well, the legend has to take some poetic license. 17:20 schwarzwald[m] If we make sure the new noise implementation doesn't have undefined integer overflow that would also fix some cases detected with the current noise code. 17:21 MTDiscord it's not like s32 exists 17:21 schwarzwald[m] Why do we use s32 instead of std::int32_t BTW? 17:21 MTDiscord it's signed 17:22 schwarzwald[m] They both are signed. 17:22 MTDiscord what is s32? ¬¬ 17:22 schwarzwald[m] std::uint32_t would be equivalent to Irrlicht's u32 type. 17:23 schwarzwald[m] I'm curious why we use Irrlicht integer types in general, when the STL provides fixed width integer types even since C++11. 17:24 rubenwardy Irrlicht was created before C++11 17:24 rubenwardy same for Minetest 17:24 rubenwardy they should replaced by STL types 17:25 schwarzwald[m] A fair enough reason! I had a phobia of using STL types because it looked like we preferred Irrlicht's. xD 17:26 MTDiscord if we use irrlichts it means we have consistency and during the rip out procedure a handful of #typedefs will fix any older code 17:26 sfan5 s16 is defined to be int16_t 17:26 sfan5 replacing any types in existing code is pointless busywork 17:31 rubenwardy oh right. s16 is better than int16_t then 17:31 rubenwardy I typedef/using int16_t to s16 in my own projects 17:32 MTDiscord schwarzwald: When Jordach says "the server" he means "his server, when run in singleplayer mode or only with a couple clients" 17:33 MTDiscord 10s of thousands of entities generate abhorrent amounts of traffic 17:33 MTDiscord and practically everything related to them is linear if not quadratic time 17:33 MTDiscord apparently Jordach's Ryzen can somewhat cope, but that doesn't mean that many entities are even remotely efficient 18:29 MTDiscord They could be better, a lot better 19:16 MTDiscord server-only builds now require the custom irrlitch right? 19:18 MTDiscord Using irrlichtmt is the safest option 19:19 MTDiscord irrlichtmt is required now 19:19 MTDiscord for server builds 19:24 MTDiscord thanks! 19:32 MTDiscord i've implemented a standalone reference version of perlin because i'm trying to find out why minetest's noise implementation is so slow 19:33 MTDiscord 80^3 noise samples should not be taking 10ms on a modern Ryzen CPU 19:33 MTDiscord it's perlin without octaves etc 19:38 MTDiscord whatever is struggling to be -O3'd i'd like to find out 19:47 MTDiscord probably a dumb question but this is without relying on disk i/o right? 20:08 MTDiscord yes 20:09 MTDiscord i'll probably hook up the reference perlin to mods that want the initial perlin paper version as a way to test if minetest's noise is actually right 22:02 rubenwardy can settingtypes.txt readable names have () in them? 23:00 ShadowNinja Regarding mapgen noise: We could look into using Simplex noise. We used to avoid it for patent reasons, but the patent apparently expired earlier this year. 23:21 Zughy[m] can anyone with a mac confirm #12507 please ? 23:21 ShadowBot https://github.com/minetest/minetest/issues/12507 -- Plus key does not detect on macOS 23:21 Zughy[m] same with #11791 23:21 ShadowBot https://github.com/minetest/minetest/issues/11791 -- Shift-click in inventory not working 23:59 MTDiscord good news, my reference implementation works flawlessly