Time Nick Message 18:41 Krock Meeting in 20 minutes 18:58 nerzhul_ oh a meeting, what are the topics Krock ? 18:58 Zughy[m]1 https://dev.minetest.net/Meetings#2021-01-23 19:01 Krock nerzhul_: mainly 5.4.0 release stuff 19:02 nerzhul_ Krock: can i add a point about irrlicht and our future with it ? We are very hesitating on it, maybe we can thing about a 6.0 release with an embedded irrlicht with our fixes, overview. As a detached library first then we will see 19:03 Krock no idea how many 5.x.0 releases there will be but yes sure 19:04 nerzhul_ we may think about a 6.0 when this new library will be used as a replacement, or maybe when we start to use our own very close irrlicht if we decide it 19:05 Krock though it will result in additional effort for package maintainers and users to build Minetest if an external repostiroy is required (non-bundled) 19:05 Krock there's no way around I guess... 19:05 v-rob Why would it require a 6.0? It only immediately affects the build process, no? 19:05 sfan5 yes, it wouldn't require a 6.0 19:06 nerzhul_ it's a rupture in our dependency system, as you like in terms of naming, for me it's a major milestone in terms of project 19:06 nerzhul_ anyway let's discuss that after release point, ok ? 19:07 Krock now that we're already discussing the SDL/Irrlicht part we can just go ahead with this 19:07 Krock it needs to be discussed sometime anyway 19:08 Krock if possible, a dual system using the official Irrlicht branch and custom branch would be amazing. is that feasible or just unnecessary effort, nerzhul_ ? 19:08 nerzhul_ it's a very huge effort, but possible, the problem is that you want SDL which requires a rupture in the irrlicht AND mt parts :p 19:08 Krock so you'll probably need a separate branch for it 19:09 sfan5 why separate branch? 19:09 sfan5 so we can return to upstream irrlicht one day? 19:09 nerzhul_ i'm not sure regarding the project lifetime it's an option haha 19:09 nerzhul_ but a branch for the first developments is require 19:09 Krock I'm not sure if SDL2 should become a dependency right now or whether all changes are applied at once 19:10 Krock latter means a separate branch which then can be merged into master once it's ready 19:10 nerzhul_ someone said SDL2 has huge overhead in terms of binary 19:10 nerzhul_ at runtime 19:11 sfan5 I think we should integrate sdl in a way that code is still theoretically compatible with normal Irrlicht 19:11 Krock would be nice to know who said that, to judge what *huge* means in this context 19:11 sfan5 so not just replace SEvent with SDL_Event, but translate SDL events into SEvent in Irrlicht's code 19:11 Krock sfan5: but according to nerzhul_ that'll cause a lot more effort 19:11 v-rob As far as I can tell, one of the biggest problems with SDL is text input. SDL2 uses real IME input while Irrlicht has a `Char` field in KeyInput. Instead of fixing this in Minetest, I would suggest making a polyfill from Irrlicht->SDL's events and not changing Minetest yet. 19:12 sfan5 Krock: well that doesn't have to literally be a seperate branch, more of a thoeretical possibility 19:12 sfan5 ...if you get what I mean by that 19:12 Krock I think I know what you mean, yes. 19:12 sfan5 v-rob: yeah we could add a wchar_t* to SEvent for that 19:13 sfan5 it will not be compatible with normal Irrlicht but it would not be far from it if there ever is need to use vanilla Irrlicht 19:13 sfan5 this would also be advantageous if we do not want to port Android to SDL2 immediately 19:13 nerzhul_ yep the keyboard handling in SDL2 is totally different 19:13 nerzhul_ then we must require our GUI parts 19:13 Krock is SDL2 a singular big blob, or can parts be imported individually, like the rendering/input part? 19:14 v-rob Trying to make IME in EditBox, the chat, etc. is likely to have errors and IMO belongs to the formspec replacement 19:14 nerzhul_ we should not use SDL2 rendering part except if you plan to replace GUI formspec with a SDL2 based other layout 19:14 v-rob I see no need for SDL's rendering API, mainly events and windowing 19:14 sfan5 agree ^ 19:14 nerzhul_ we are close to that 19:15 Krock right. think of it as an example then. we won't need networking either 19:16 Krock maybe that's just be being too greedy 19:17 Calinou SDL rendering is pretty oudated, it's not hardware accelerated 19:17 Krock so initially, development on a separate Minetest branch, and merge once it's usable? Is that a plan, nerzhul_ ? 19:17 pgimeno Calinou: I presume we are talking about SDL-GL 19:17 v-rob Well, SDL_Surface is not hardware accllerated, but SDL_Texture is 19:18 pgimeno so the idea would be to keep Irrlicht for the scene graph and rendering? I have issues with that 19:18 Krock Irrlicht is an issue and the goal is to incrementally replace it 19:18 pgimeno I think it should eventually be replaced, and SDL is not a bad choice 19:18 nerzhul_ irrlicht jsut take the SDL_surface and does its opengl stuff on it, look at my branch 19:18 Krock user input just happens to be the first use-case 19:19 rubenwardy Calinou: we'd never be able to use SDL rendering, it's 2d 19:19 rubenwardy We're talking about its ability as an input library and window manager 19:19 Krock nerzhul_: so you added SDL2 as Irrlicht device which is then used for rendering, or am I seeing that wrong? 19:19 v-rob The only way to use SDL with 3D is through OpenGL, Vulkan, or Direct3d 19:20 nerzhul_ yes it's a thing like this 19:20 nerzhul_ and sdl2 is used for text input but it's not clear yet on the MT side 19:21 pgimeno if migrating allows dead keys (accents) to be typed, that would be awesome 19:22 v-rob It would eventually, but likely not immediately. It would also allow things like Chinese and other such languages as well 19:23 Krock okay. noted that. I'll update the entire Meeting wiki section after the other points are done 19:23 Krock just to have a rough overview of what's going to happen next 19:23 v-rob Along with improved joystick/game controller and touchscreen events 19:25 Krock okay. Starting with the 5.4.0 release topic 19:25 Krock =============================== 19:25 Krock all issues and PRs which should go into 5.4.0 should be marked by now 19:26 Krock according to rubenwardy (I believe), the desired freature freeze is today or in a week? 19:26 rubenwardy I reckon so 19:27 Krock other thoughts? I'd be okay with a freeze from today but in exchange the freeze might last longer for difficult issues 19:28 Krock Alpha textures now have a fix. the ContentDB issue for Windows should be easy to fix, given that someone can test it 19:29 Krock my concern is the get_player_information where no clear information about 5.4.0-dev is available 19:30 Krock from what I understand the problem is that the callbacks are run even after the player is kicked 19:31 sfan5 that's avoidable in mods 19:31 Krock i.e. mod issue and does not need fixing? 19:31 sfan5 the issue remains that the peer can disappear at any time before the player disappears 19:31 sfan5 or at least that's what I guess, haven't looked into this deeply 19:32 Krock so it's an async issue while unloading a player's data 19:32 sfan5 I have no idea about #9352 yet 19:32 ShadowBot https://github.com/minetest/minetest/issues/9352 -- minetest.get_player_information sometimes return nil (running at mod load, in singleplayer) 19:32 rubenwardy -> mod load 19:32 rubenwardy the singleplayer doesn't join until after mods load, right? 19:32 rubenwardy I should probably read more than the title 19:33 Krock okay. It would be nice to somehow reproduce this error reliably 19:33 sfan5 the title is wrong yea 19:33 numzero I’m working on using SDL for input now 19:33 numzero text input works 19:33 Krock numzero: thanks for the input. will note that. 19:34 numzero touch input almost works 19:34 v-rob Wait, are there two people working on SDL at the same time? 19:34 rubenwardy pun alert 19:34 Krock rubenwardy: intended 19:34 numzero I started my work off nerzhul’s branch 19:35 nerzhul_ numzero you have PR on my branches ? can be nice 19:36 numzero I have two branches now, one compatible with nerzhul’s (so he can pick commits from that) and one merged with ogl-es (Irrlicht 1.9) 19:36 Krock I'll also have a look at the player information issue but recently my time is also kinda limited 19:37 Krock let's see 19:38 Krock #10636 19:38 ShadowBot https://github.com/minetest/minetest/issues/10636 -- Improve touch event handling by numberZero 19:38 numzero as using SDL with Irrlicht 1.8 is kinda pointless, it doesn’t support all our target platforms 19:38 numzero so there will be need to use other Irrlicht version anyway 19:38 Krock Q: how many people already tested this? 19:38 Krock or rather, what's the state of the PR? mergeable? 19:39 rubenwardy I was going to, but then my computer ran out of storage and I got distracted on house keeping 19:39 rubenwardy also text wrapping 19:39 numzero Krock: depends on #10808 19:39 ShadowBot https://github.com/minetest/minetest/issues/10808 -- Improve irr_ptr by numberZero 19:39 rubenwardy implementing working word wrapping took much much longer than expected 19:39 numzero which has line length issues 19:39 Krock numzero: yes I saw that. shouldn't be a major concern. 19:39 Krock // clang-format off 19:39 Krock // clang-format on 19:40 Krock fixed 19:40 numzero which is possible to “fix”, or agree to #10792 19:40 ShadowBot https://github.com/minetest/minetest/issues/10792 -- Clang-format: increase column limit from 90 to 105 by Zughy 19:40 Krock how about Gradle? is that fixed in master? 19:41 Krock that's the other of the two failed builds/checks 19:41 sfan5 yes 19:41 Krock good. thanks. 19:42 rubenwardy TIL operator== is not commutative 19:43 Krock rubenwardy: but you'd have time in the next few weeks to test the input PR? 19:43 rubenwardy yeah 19:43 Krock thx 19:44 Krock #/10819 19:44 Krock #10819 19:44 ShadowBot https://github.com/minetest/minetest/issues/10819 -- [DONT SQUASH] Rework use_texture_alpha by sfan5 19:44 Krock I was going to test this like yesterday. Will give it a try again today by testing all registered nodes in a few games 19:45 Krock also backwards compat needs testing, which is unlikely covered by Lone_Wolf's review 19:46 Krock backwards compat from an old server -> new client and new server -> old client 19:47 Krock #10555 19:47 ShadowBot https://github.com/minetest/minetest/issues/10555 -- Mapgen settings in a game's minetest.conf get ignored 19:48 Krock does this have to be fixed for 5.4.0? IMO the only proper solution would be to once and for all clean up the settings structure in 310661 19:49 Krock #10661 19:49 ShadowBot https://github.com/minetest/minetest/issues/10661 -- [MANUAL SQUASH] Settings: Proper priority hierarchy by SmallJoker 19:49 Krock whereas I tested many cases, it is still a deep change and would like to get some feedback there in case the issue has priority 19:50 Krock also waiting for Casimir's answer to be sure 19:50 sfan5 settings stuff should be fixed for 5.4 IMO 19:52 Krock okay. I'll add the milestone to that PR because I don't think any other hacky method would work reliably due to overwriting issues 19:53 Krock so when's the feature freeze starting? give it some more time to find a solution for get_player_information, or put on a bit of pressure by starting it today? 19:53 Krock opinions? 19:54 sfan5 are there any feature PRs ready enough to get in? 19:54 sfan5 if not it doesn't really matter when the freeze starts 19:54 rubenwardy v-rob has a PR with approval 19:54 rubenwardy builtin translate is also close to being approved 19:55 Krock https://github.com/minetest/minetest/pulls?q=is%3Apr+is%3Aopen+label%3AFeature+label%3A%22One+approval%22 19:55 Krock translated builtin would be indeed good to have 19:55 Krock rubenwardy: are you talking about #10083 19:55 ShadowBot https://github.com/minetest/minetest/issues/10083 -- FormSpec: Add list spacing, slot size, and noclip by v-rob 19:56 rubenwardy yeah 19:56 rubenwardy it has approval, v-rob can merge if he's still happy with it 19:56 rubenwardy not important for 5.4.0 though, but well over due 19:56 v-rob Yeah, I'm still happy with it. 19:57 Krock looks good from a quick glance 19:57 v-rob PRs for it have been around for about three years now 19:57 Krock so merge that and start feature freeze today? 19:58 sfan5 I vote for starting it next week 19:58 Krock I also feel safer that way 19:58 rubenwardy fine by me 20:01 kilbith before you start to freeze, I remind you that I need that too: https://github.com/minetest/minetest/pull/10265 20:01 Krock !title 20:01 ShadowBot FormSpec: 9-slice images, animated_images, and fgimg_middle by v-rob · Pull Request #10265 · minetest/minetest · GitHub 20:01 rubenwardy I posted review comments for that, yet to be addressed 20:01 v-rob I can fix those today 20:01 kilbith thank you 20:03 Krock game#2803 20:03 ShadowBot https://github.com/minetest/minetest_game/issues/2803 -- Lossless PNG Compression by ExeVirus 20:04 Krock rubenwardy: what would you like to discuss? 20:04 rubenwardy What the issue is that prevents from doing this optimisation 20:04 Krock reading.... 20:05 rubenwardy If it's just the leaves thing, then those have been whitelisted by the author to retain RGB when a=0 20:05 Krock I've seen that issue before 20:05 Krock game#2801 20:05 sfan5 the mipmapping done by the engine code(?) doesn't ignore transparent pixels 20:05 ShadowBot https://github.com/minetest/minetest_game/issues/2801 -- Foliage using allfaces_optional becomes too dark with distance (most leaf texture transparent pixels need improved colouring) 20:06 rubenwardy oh interesting, so you either break mipmapping or you break opaque leaves 20:07 Krock opaque leaves are a sin 20:07 sfan5 hence the transparent pixels need to still have colors 20:07 rubenwardy for Android I can seem them being worth it 20:08 v-rob I can live with opaque leaves; it's opaque water that I can't stand 20:09 Krock so either whitelist or compress the textures in a way without discarding the alpha colors 20:10 Krock I think the opaque leaves issue is less important than bad mipmapping 20:10 Krock IIRC that opaque mode is done using a certain color threshold 20:11 sfan5 or fix the code 20:11 Krock imageCleanTransparent imagefilters.cpp 20:11 sfan5 but I don't know if the engine does the downscaling itself 20:11 rubenwardy that PR already whitelists to preserve alpha colors 20:13 Krock "The table of contents is too big for display. " thanks GitHub 20:14 Krock fixing the opaque mode then seems the best solution to me, in combination with whitelisted textures 20:15 Krock why whitelist? those textures that have alpha contents do show it most likely to the client 20:15 Krock exceptions are overlay textures 20:15 rubenwardy whitelist meaning that the optimiser doesn't remove the colors 20:16 Krock it's a matter of a few bytes whether there's color in alpha=0 pixels. not optimizing any texture using that method would also be a solution 20:17 rubenwardy exe_virus: thoughts? 20:17 MTDiscord @exe_virus 20:18 MTDiscord I already whitelisted those textures 20:18 MTDiscord There won't be any issues 20:18 MTDiscord It totals an extra like 200 bytes or so 20:19 MTDiscord The PR contains those whitelists. Though I want to go over those 30 one more time before it's officially merged, since I did that for the rest of the textures in mtg 20:20 MTDiscord You mean paramat game? 20:20 rubenwardy exe: the suggestion is to not do it for all textures, to avoid the whitelist. Does that have a big effect? 20:21 rubenwardy well, I suppose it's unlikely we'll be adding more leaves given MTG is in feature freeze 20:21 MTDiscord All textures? Yeah it has a larger effect when we don't use 8-bit indexed for all textures. 20:21 MTDiscord But this is only an issue for allfaces drawtypes 20:22 MTDiscord Hence the 30 that were reverted 20:22 Krock no, also an issue for plantlike 20:22 Calinou are opaque leaves worth supporting when we have simple (glasslike) leaves with denser textures? 20:22 Krock also glasslike and nodeboxes (glass) 20:23 Krock without alpha=0 colors those will also look bad from a distance 20:23 Krock but I doubt those have such colors already 20:23 MTDiscord No worries, the original PR in 2016 that originally reverted the textures is where I got the 30 files. I didn't question it but it's more than the 15 or so that should be allfaces. If you want I can go back and revert any drawtypes you want, just give me the list 20:24 Krock it's just what came into my mind just now. I cannot reliably tell where exactly it's needed 20:24 MTDiscord So right now it's: Plantlike Allfaces+ Glasslike Glass-nodeboxes 20:24 MTDiscord Gotcha am I looking for anything with transparency? 20:25 Krock textures where the transparency is shown to the client, and where the placed node looks bad from a distance 20:25 Krock sorry, I cannot give a better explanation than this 20:26 Krock (note: mipmapping must be enabled) 20:26 MTDiscord Yeah, and you can't have a nvidia gpu 20:26 MTDiscord Only amd or intel integrated with mipmapping has the issue 20:27 Krock can reproduce it on a GTX 1050 20:28 Krock the meeting is over, I presume. 20:28 Krock I don't have any additional points on the list and the wiki page is updated now 20:30 MTDiscord Okay, I'll research what happens if we don't use 8-bit indexed mode for the pngs 20:31 kilbith Krock, well you forgot to mention inventory slot spacing but *sigh* 20:32 Krock oh sorry. I thought we were talking about the badly scaled mipmap textures. 8-bit indexed shouldn't have an influence there 20:32 Krock not sure though 20:32 Krock kilbith: I need to move the section so I'll add that now. thanks for pointing out 20:34 kilbith lhof would have talked about ZSTD in this meeting, too bad 20:35 MTDiscord We should have 20:35 MTDiscord I am clamoring for ZSTD 20:35 MTDiscord But I'm not a core dev haha 20:41 v-rob Any problems with me merging #10083? 20:41 ShadowBot https://github.com/minetest/minetest/issues/10083 -- FormSpec: Add list spacing, slot size, and noclip by v-rob 20:45 kilbith author being core-dev equals to a pre-approval 20:45 v-rob Just making sure no one had any issues with it 20:47 kilbith btw I noticed that hovering a hypertext[] with no scrollbar inside a scroll_container, blocks the container scrolling 20:48 kilbith solution: disable the scrolling of hypertext without scrollbar 20:48 v-rob Sounds like hypertext is eating the mouse event 20:48 kilbith yes 20:49 kilbith it should happen when there's actually a scrollbar when the content exceeds the limits 20:56 ronoaldo Hey! I have tested the PR from drag and drop improvements on both desktop and mobile with no issues, looks good in terms of not adding side effects on my phone, and fixes the drag on mobile 21:20 MTDiscord Okay, @rubenwardy I closed the old mtg compression PR and opened a new one #2808. It only uses optipng.exe and does not change any bit depth settings, only strips metadata and attempts maximum zlib compression. Result still saves 93KB. Original PR only saved 11.2KB more, so not a big deal. Low hanging fruit first. 21:20 ShadowBot https://github.com/minetest/minetest/issues/2808 -- Add MoveSomewhere inventory action by est31 21:24 rubenwardy merging #10838 in 10 21:24 ShadowBot https://github.com/minetest/minetest/issues/10838 -- Do not close game on escape in error windows by yamanq 21:33 Zughy[m]1 #10792 comments addressed 21:33 ShadowBot https://github.com/minetest/minetest/issues/10792 -- Clang-format: increase column limit from 90 to 105 by Zughy 21:37 sfan5 do we know if clang-format still runs with those settings 21:40 rubenwardy yes it does 21:41 sfan5 there's no CI run in the PR 21:41 rubenwardy the workflow will be broken then 21:42 rubenwardy .clang-format needs to be added to both paths in .github/workflows/cpp_lint.yml 21:43 v-rob Addressed issues in #10265 21:43 ShadowBot https://github.com/minetest/minetest/issues/10265 -- FormSpec: 9-slice images, animated_images, and fgimg_middle by v-rob 21:43 rubenwardy thanks, I'll take a look later 21:46 kilbith v-rob, we can't set a spacing of 0 for slots of size < 1? 21:48 kilbith v-rob, https://i.imgur.com/uAwCDQH.png 21:49 kilbith size = 0.7; spacing = 0 21:52 v-rob Uhh, let me check that out 21:57 v-rob Oh gee, you're right. I did some calculations with imgsize instead of slot_scale. I will fix that right away. 22:21 kilbith v-rob, so? 22:22 v-rob Just compiled, making the PR now 22:26 kilbith testing 22:28 kilbith ok works as expected 22:35 rubenwardy v-rob: FYI: you can merge trivial bug fixes by yourself, you just need to announce in this channel with notice 22:37 v-rob ok 22:38 rubenwardy it has to be trivial (=small and unlikely to cause a problem) and has to be a bug fix 22:38 v-rob Right 22:39 v-rob Then, merging #10860 and #10861 in 10 minutes 22:39 ShadowBot https://github.com/minetest/minetest/issues/10860 -- Make hypertext and textarea have proper scroll event propagation by v-rob 22:39 ShadowBot https://github.com/minetest/minetest/issues/10861 -- Fix formspec list spacing by v-rob 23:49 kilbith_ okay there's another bug with list styling 23:50 kilbith_ v-rob? 23:57 kilbith #10862 23:57 ShadowBot https://github.com/minetest/minetest/issues/10862 -- Slot size of >1 ignores height size