Time Nick Message 00:04 [MTMatrix] Another reason to back my self-assign proposal and extend it for whatever PR really: https://github.com/minetest/minetest/pull/13577#issuecomment-1894699536 09:54 sfan5 in this case we shouldn't ask for PRs to be rebased unless someone has ever looked at them 11:07 sfan5 anyone with macOS around who could test SDL2 with latest irrlicht? 14:04 MTDiscord Will lua on each mapgen thread go into 5.9 even if it comes before fossdem? 14:26 rubenwardy Are there any features that deserve a rush to release? 14:26 rubenwardy Sdl is still bugged and GLTF isn't in, I'd not rush to release 14:52 [MTMatrix] I agree with ruben, it's risky and it'd put pressure on development 14:53 [MTMatrix] unless a reasonable amount of core devs have enough time in these two weeks to cover potential issues that might rise 14:53 [MTMatrix] (luatic said that GLTF should cause no big issues though. Let's just mark it as experimental and the problem is done, innit?) 15:02 MTDiscord I do not want a repeat of what happened with dynamic shadows. 15:03 MTDiscord I know a lot of people have expressed excitement about glTF support, but I do not think that is a good reason to rush it. 15:11 Desour if we make a release with easy to fix obvious bugs, that might lure new developers. otherwise idk why you'd want to release again after under 2 months. if you want to showcase new shiny features, take the other features we aggregated in recent years, and show them by making new mods. or even better, find ways to showcase what makes minetest actually interesting 15:15 celeron55 in the case where no release is rushed, the upcoming features can of course still act as conversation points at fosdem 15:48 [MTMatrix] re Desour "find ways to showcase etc.": the only thing I can think of for develoeprs is CDB and scripting in Lua. Because, I mean, both GUI and HUD suck, we're still stuck on OpenGL 2 and with an engine forgotten by humanity, there is no way to go from a server to another, we've still got plenty of things to dehardcode 15:48 rubenwardy Minetest: Jank as a service 15:49 rubenwardy ~~GUI replacement for 5.9?~~ 15:49 Desour still, there are so many contributors in minetest, so something must be drawing us in, even if it's the suck-ness of minetest, and therefore resulting fun of being able to improve things everywhere 15:50 [MTMatrix] for me is showing the middle finger to Microsoft whilst doing something I like, and I won't even hide it. We're racing against the most sold videogame in history, fuck them 15:50 [MTMatrix] But I can't go to people and say "hey, do you want to use that lovely middle finger of yours? ;)))" 15:51 Desour and from the users perspective there must also be interesting things, otherwise they wouldn't be here. maybe it's nice that you can do so much hacky stuff with out API, or that you can host servers where arbitrary content is sent to the client 15:51 Desour haha 15:52 [MTMatrix] oh no, trust me, I despise hacky stuff (e.g. using invisible entities to immobilise players) 15:52 Desour it's FOSSdem, so showing how we encourage free licensing on forums and contentdb is also noteworthy. when contentdb migration btw? 15:53 Desour "I despise hacky stuff" so stockholm didn't kick in yet for you? 15:53 [MTMatrix] It's love and hate 15:54 rubenwardy "when contentdb migration btw" what do you mean? 15:54 Desour oops, s/contentdb/codeberg/ 15:54 [MTMatrix] anyway: it'd be nice if Minetest showcased featured servers as well. Bluntly pushing for mine and CTF, can't tell about others 15:55 [MTMatrix] Desour: oh I'm in haha 15:58 Wuzzy ive tried multiple times to tacke #12578 but i struggle to find the cause of the bug. i know it has something to do with the shader but i fail to make any progress 15:58 ShadowBot https://github.com/minetest/minetest/issues/12578 -- Texture overlay color of items is different as wielditem than in inventory/dropped item 15:59 Desour "showcased featured servers" <- when I decide if I want to play a mostly-multiplayer game, it's super important that there's actually active servers and players. so this is definitely a good idea. (I didn't play much besides devtest recently though. could push for your land though) 16:01 Wuzzy about "showcased featured servers": Instant suggestion for Inside The Box 16:01 Wuzzy about bug 12578 again... i have stared at WieldMeshSceneNode::setItem and this line: u32 shader_id = shdrsrc->getShader("object_shader", TILE_MATERIAL_BASIC, NDT_NORMAL); 16:03 Wuzzy so whats weird the same code is also used in client/content_cao.cpp but for this, the coloriing is correct o_O 16:04 Wuzzy whats doubly weird is not just that the wieldmesh is not only slightly brighter but the hue seems to be slightly off, too 16:06 Desour tested. the difference is huge with the color in the reproduction steps 16:06 Desour red-ish vs white-yellow 16:06 Wuzzy yeah i chose a color where difference is strongest but other cases are more sublte. still, its always a noticable difference 16:07 Wuzzy i guess i just need to actually start learning OpenGL to even have a chance of fixing this ;-) 16:09 Wuzzy buuuut, i have a suspicion somewhere in the code the color change and/or shading is applied ... twice? but why does the *hue* change too then? 16:19 Desour the object vertex and fragment shader both apply daylight ratio stuff. if I comment out the one in the vertex shader, it's fixed 16:20 Desour (see client/shaders/object_shader/opengl_vertex.glsl:125) 16:21 Desour it's kinda weird that the vertex color is shaded at all 16:22 Desour (and it's so old-style that we're using vertex colors for this feature. a uniform would make more sense) 16:31 Desour merging #11073 in 10 min 16:31 ShadowBot https://github.com/minetest/minetest/issues/11073 -- Add rotation support for wallmounted nodes in 'ceiling' or 'floor' mode by Wuzzy2 16:50 sfan5 new: #14268 16:50 ShadowBot https://github.com/minetest/minetest/issues/14268 -- Use newer IrrlichtMt by sfan5 17:01 [MTMatrix] sfan5: are you trying to release a 5.9 before FOSDEM? 👀 17:08 sfan5 not specifically, right now I'm just trying to get some outstanding things on irrlichtmt solved 17:37 sfan5 looking at the blockers we have that's probably also not too realistic 17:46 sfan5 testing welcome btw 17:47 sfan5 @luatic can I merge #14225? 17:47 ShadowBot https://github.com/minetest/minetest/issues/14225 -- Add bulk_get_node function by sfence 17:48 Krock why do we need this again? the (assumed) relative positions to a node (whose surroinging nodes are interesting) have to be calculated manually anyway 17:48 Krock so this only saves some Lua -> C calls, or is there more to it? 17:49 Krock I'll have a look at the other PRs to give them at test. feel free to let me know about such that have priority for you 17:49 sfan5 consider that a 3x3x3 cube needs 27 get_node calls otherwise 17:50 sfan5 so just getting a nodes direct neighbors is relatively expensive 17:51 sfan5 #13616 could use a revierw 17:51 sfan5 -r 17:51 ShadowBot https://github.com/minetest/minetest/issues/13616 -- Meshgen tests by numberZero 17:51 Krock I was thinking about providing fixes rule tables directly, for relative positions. In some cases that might be easier to use - then again it's less generic than an absolute node lookup function 17:51 Krock s/fixes/fixed/ 17:52 sfan5 and #14256 could use someone with an opinion on `IrrlichtErrorCapturer` 17:52 ShadowBot https://github.com/minetest/minetest/issues/14256 -- Return to the main menu if a shader compilation fails by HybridDog 17:52 sfan5 I think performing the vector additions in Lua won't be the bottleneck 17:52 sfan5 maybe more specialized APIs could make sense in the future 17:53 sfan5 but I think the "get_node > bulk_get_node > VoxelManip" progression makes a lot of sense 17:53 Krock fair point 17:53 sfan5 anyway afk 30m and I'll merge some of the one approval PRs after that 17:54 Krock hmm... I wonder how badly the VManip could possibly perform 17:56 Desour the bulk node get function could probably be optimized more if the node conversion was done in lua. afaik pushnode as of now calls a lua function, but we could instead return an array of lua numbers, and call pushnode in lua, for less C->lua calls. anyways, this can happen later, having a bulk node get function api in place first is probably fine 18:39 MTDiscord I think a "poor" API may limit us performance-wise; it'd be better to start with an API that is designed to be efficient right away - a 1.3x performance improvement doesn't seem like a whole lot. 18:39 MTDiscord But I'll have to benchmark, could very well be that my suggestions aren't as useful as I think they are. Also, while 5.9 is still in dev, I believe we'd be free to change the API? 18:40 sfan5 practically yes but we try to avoid it anyway 18:41 MTDiscord I'll try to make some benchmarks soon (TM) 18:42 sfan5 fwiw I think it's worth not deviating from the {name=..., param=1..., param2=...} and vector table structure even if it costs performance 18:42 sfan5 for an api named [gs]et_node that is 18:44 Krock perhaps we could speed up the API a bit by getting rid of the env indirection 18:48 Krock nvm. there's not much to optimize. 18:54 sfan5 merging #14265, #14255, #14217, #14090, #13880 in 10m 18:54 ShadowBot https://github.com/minetest/minetest/issues/14265 -- [no squash] More maintenance stuff by sfan5 18:54 ShadowBot https://github.com/minetest/minetest/issues/14255 -- [no squash] Avoid use of select(2) by sfan5 18:54 ShadowBot https://github.com/minetest/minetest/issues/14217 -- [no squash] Networking improvements / hardening by sfan5 18:54 ShadowBot https://github.com/minetest/minetest/issues/14090 -- Fix out of range enum casts in deSerialize functions by cx384 18:54 ShadowBot https://github.com/minetest/minetest/issues/13880 -- Use performant and safe data structure for active objects by sfan5 18:56 Krock I remember there were some sha256 fixes from our side - years ago. Updating the file should be fine, though. 19:07 sfan5 hopefully nothing breaks now 19:15 nrz i see it's lib update time 😄 19:34 Krock added some more benchmark results to https://github.com/minetest/minetest/pull/14225#issuecomment-1896459828 in case anyone wonders 19:39 sfan5 if anyone is looking for a free puzzle try figuring out why these tests fail https://github.com/minetest/minetest/actions/runs/7560583285/job/20586929096#step:7:271 19:40 sfan5 (warning: windows) 19:40 Krock oh I've seen this one before 19:40 Krock problem being I don't remember the rewason 19:41 Krock possibly one of my first PRs were something in that direction 19:50 Krock affected bytes: 00 00 d1 1e 19:57 Krock 53.53400039672852 19:57 Krock https://pastebin.com/raw/3kXdg4A9 19:58 Krock adding some tolerance (0.0001f) should fix it 20:11 MTDiscord sfan5: agreed, but we can provide a get_node_fast API or similar that returns a name, param1, param2 20:13 Wuzzy what's the difference between minetest.after(0, my_function) and minetest.register_on_mods_loaded(my_function)? (assuming i run this code at loading time, i.e. not nested within another function) 20:14 sfan5 before the first server step the environment isn't initialized 20:14 sfan5 and on_mods_loaded probably runs before env init 20:16 Krock Wuzzy: latter still allows you to register nodes, aliases, abms and decorations 20:17 Krock afterwards the NodeResolver code is run, which makes any manipulation useless. 20:18 Wuzzy so minetest.after(0, ...) is basically "after the first server step"? 20:19 celeron55 on_mods_loaded runs literally immediately after every mod's init has been run, nothing happens in between 20:19 Krock minetest.after is run by the server globalstep (Lua side) 20:20 Krock so you can either use minetest.after or do your own timings in a globalstep callback. it's the same thing. 20:20 Wuzzy ah ok =) 21:38 grorp luatic: can we please get rid of the Github-specific markdown "alert" syntax introduced in https://github.com/minetest/minetest/commit/a8cf10b0b523bf4c3234be29b0dd51e719cb00c1? it doesn't render nicely on api.minetest.net and it's inconsistent with the rest of the API documentation. 21:40 grorp also, where was the "pushing in 5 min" message for that commit, as required by the Git Guidelines? 21:41 grorp (the question mark wasn't supposed to be part of the URL) 21:44 MTDiscord grorp: Is the "pushing in 5 min" announcement really useful for tiny doc edits? 21:46 grorp I guess usually not, but in this case I actually have something to complain about, so it would have been useful :) 21:46 MTDiscord grorp: True, sorry 21:46 MTDiscord I suppose I'll piggy-back the next doc / commit I make to fix this. Making a separate commit for this would seem excessive. 21:46 MTDiscord PR / commit* 21:47 ROllerozxa "it doesn't render nicely on api.minetest.net" well, what if we made it render nicely :) 21:49 celeron55 if a "gotcha!" syntax like that gets foothold, soon the docs are going to be absolutely littered by it 8) well maybe not 21:50 MTDiscord I think such a "gotcha!" syntax is nice, though GH's extension in particular may not be. Slightly sucks that Markdown isn't AsciiDoc here. 21:50 MTDiscord It depends on whether we actually attempt to document every gotcha or whether it's more efficient to simply list exceptional cases where there isn't a gotcha. 21:51 MTDiscord Suggestion: Document every second gotcha, then document the gotcha that only every second gotcha is documented. 21:51 celeron55 sounds perfect 21:51 ROllerozxa mkdocs (or the markdown library it uses) has a markdown extension system which I assume makes it possible to shove some python code somewhere to add custom extensions to the parsing output. maybe something to look into 21:52 MTDiscord The old documentation was a text file; it was not GitHub specific. Changing to markdown helped readability a lot, but I like to view this in my local web browser. 21:53 ROllerozxa it's not the end of the world though, the GFM alert syntax degrades gracefully into a blockquote with a warning header... 21:53 MTDiscord Gotcha: this gotcha left intentionally undocumented. 21:54 MTDiscord Used to be even nicer (bold note text in a blockquote) but they removed that syntax. 22:01 grorp I'm not sure whether a "colorful boxes" syntax is going to result in better or worse docs. You'd have to make sure that enough continuous, non-boxed text remains to keep the docs readable. 22:23 lhofhansl sfan5: see #14271 (#13880 causes a seg fault) 22:23 ShadowBot https://github.com/minetest/minetest/issues/14271 -- SEGV (on Linux) with latest Minetest Master 22:23 ShadowBot https://github.com/minetest/minetest/issues/13880 -- Use performant and safe data structure for active objects by sfan5 22:26 MTDiscord I've considered opening an issue to ask for the creation of a minetest.segfault() API, since I need to be able to reliably trigger one to test my segfault trapping measures, but I was going to wait until it looked like there might be a shortage first. 22:47 lhofhansl As a Lua API? Or triggered from C++? 23:21 MTDiscord In my case I want a /crash chat command, so it'd need to be at least available to Lua somehow.