Time Nick Message 00:29 lhofhansl Should I be able to take the APKs generated by our Github build workflows and install them on (say) a Pixel 8? 00:30 lhofhansl I can install other downloaded APKs, but the Minetest ones fail to install. Has anybody tried that? (I tried both of the ARM flavors) 00:38 rubenwardy they're unsigned 00:38 rubenwardy you need sign them first 02:25 sfan5 tip: .../android-sdk-linux/build-tools/30.0.3/apksigner sign --ks ~/.android/debug.keystore *.apk 02:30 sfan5 ah and the password for it is 'android' 05:50 lhofhansl rubenwardy: I see. Thanks. 14:22 sfan5 merging #14220, #14207 in 5 to 10 m 14:22 ShadowBot https://github.com/minetest/minetest/issues/14220 -- Add "--needed" to Arch command to avoid reinstalling packages by Sovenok-Hacker 14:22 ShadowBot https://github.com/minetest/minetest/issues/14207 -- [no squash] Windows stuff by sfan5 14:22 sfan5 (m = minutes) 14:27 Krock wouldn't it be more intuitive to have --needed in front? 14:28 Krock if that's an option 14:29 sfan5 can do 14:31 Krock I hope it works 🤞 14:31 Krock thanks for the change 14:32 MTDiscord by the way, would core devs mind backporting Lua 5.2 features to our vendored Lua 5.1? 14:33 MTDiscord the goal would be to remove the difference between LuaJIT / PUC builds. currently modders can write code that relies on JIT features (like goto) and hence won't run on PUC builds. 14:33 Krock by the way would it be possible to somehow add our MT modifications into an official Lua package? this way we could offload the maintenance burden 14:34 Krock your points are valid, though. would be nice if there was already such library out there. we cannot possibly be the only ones in this situation 14:34 Krock unless we'd entirely switch to LuaJIT? 14:37 sfan5 I am against maintaining a "real" Lua fork 14:37 sfan5 "real" = feature backports or so; not just some security / fix patches 14:38 sfan5 if you want lua 5.2 features the best approach is to revive the PR we had that added 5.2 support 14:38 sfan5 though I don't know if that then results in the opposite problem where code works on 5.2 but not luajit 14:39 MTDiscord sfan5: it would, 5.2 replaces setfenv with the weaker _ENV for example 14:39 sfan5 then that's off the table 14:39 MTDiscord ok 14:39 sfan5 maybe we should be looking at embedding JS? ;) 14:39 MTDiscord oh no 14:40 MTDiscord may I suggest our own lisp dialect instead ;) 14:40 sfan5 there are actually some good embeddable js libraries, not V8 14:41 Krock or switch to Chromium Embedded Framework. Features: JS scripting, formspec rendering, HTML5 + CSS3 support 14:41 MTDiscord those embeddable libraries probably don't achieve V8 / LuaJIT performance though 14:41 Krock needless to say that's about 20 times the entire Minetest source code 14:42 MTDiscord the gltf PR will probably end up somewhere at 2k+ loc btw. more than half of that will be a custom gltf loader, tinygltf is a hot mess. 14:42 MTDiscord (speaking of bloat) 14:43 Krock what were again the reasons against Lua 5.4? I've heard there were quite some improvements in terms of performances compared to older versions 14:43 MTDiscord Lua 5.4 is not compatible with LuaJIT for one 14:43 MTDiscord (LuaJIT lacks 5.4 features like integers) 14:44 MTDiscord also 5.4 features like integers are likely to break mods 14:44 Krock I mentioned this to check whether LuaJIT could be superseded 14:44 MTDiscord Lua 5.4 has quite some improvements, but it's still not as good as JIT 14:44 MTDiscord performance-wise 14:44 Krock alright 14:44 MTDiscord well, I can live with 5.1 14:45 MTDiscord "somewhere around 5.1 / JIT" is probably the most popular Lua version 14:46 MTDiscord I believe we could remove 5.2+ features from JIT btw just by setting some defines, so stuff like goto wouldn't work anymore. Not sure whether that's a good idea, some modders might want to write JIT-only mods (like jitprofiler). 14:46 MTDiscord Quickjs ng and rmlui 14:46 MTDiscord Would give you chromium much smaller 14:48 MTDiscord if i interpret that table correctly, it's 30x slower than V8. that's awfully slow. 14:48 MTDiscord swapping LuaJIT for that would be like swapping a racing horse for a disabled mule. 14:49 nrz Krock it's CEL you search. Like skyrim 14:50 MTDiscord CEL? 14:50 MTDiscord Locking JIT-only features behind a barrier like the way we do insecure environment might be a decent compromise, letting us have jitprofiler while still discouraging modders strongly from locking us in 14:50 nrz CEF, it's a good idea for a major but badly documented 14:51 MTDiscord major release? 14:51 MTDiscord computer science major? 14:51 MTDiscord Chromium embedded framework I think 14:51 Krock nrz: my mention of CEF was not to be taken seriously. unfortunately I forgot to mark that 14:53 nrz Remember that switch scripting engine means big break, but we could rework pour bindings at least on callbacks to have generic plugable engine 14:56 Krock at this point it's nonsense to even try to build up a new scripting language. Just look at lua_api.md which has grown to insane lengths already. In the long run it's probably best to use a maintained version of Lua 5.1 if we don't want to make LuaJIT a hard dependency 16:06 nrz 5.1 is Not maintained no. 16:07 nrz ? 18:14 zarac anyone know if sfence / appgurueu hang around here? (got a question / comment regarding this: https://github.com/minetest/minetest/pull/14129/commits/1badd1dd60fda8d1ef4417b3ad6f46ca69d57abf ) 18:16 zarac It'd be nice if it reconnects back to original host upon transfer failure. It's not obvious to me from the commit that this is done. 18:16 zarac (so i think it doesn't) : ) .. seems like it maybe silently fails? 18:28 MTDiscord I'm appgurueu and I hang around here, but I'm not a PR author, just a voluntary reviewer :) 18:28 MTDiscord Why not just voice your concern on GitHub? sfence can address it there.