Time Nick Message 01:08 [MatrxMT] Figured out why the last two pages of https://forum.luanti.org/viewtopic.php?t=29923 don't load when I'm logged in. Stop forum spam plugin chokes on a user's email address that has a / in the local-part (before the @). 01:09 [MatrxMT] kind of annoying, also it creates a race condition where I check the thread in another container tab then go to front page to mark all forums read 08:49 sfan5 you could try pinging c55 to patch the bug 08:50 [MatrxMT] celeron55 add it to the backlog along https://forum.luanti.org/viewtopic.php?p=452107#p452107with reporting and password resets tyvm 08:51 [MatrxMT] oops here's the clickable URL https://forum.luanti.org/viewtopic.php?p=452107 08:53 sfan5 next suggestion: have a patch ready to fix it 08:56 [MatrxMT] :\ 08:56 sfan5 quickest fix is probably to skip the banhammer code in the affected template if an / is present 09:35 celeron55 maybe i'll spin up claude to patch it 09:59 celeron55 eh, this is not working 13:10 MinetestBot 02[git] 04sfan5 -> 03luanti-org/minetestmapper: Add bounds checks to block decoding code 130d40f9d https://github.com/luanti-org/minetestmapper/commit/0d40f9d47ba8a733f269d182d1ac1310ceb9fd62 (152026-08-08T13:07:37Z) 13:27 farribeiro|irccl Good morning, I wanted to ask about Luajit, which received some new features in the last commit. https://github.com/LuaJIT/LuaJIT/issues/1475 13:28 MTDiscord To ask what? 13:28 MTDiscord Don't ask to ask, just ask 13:28 farribeiro|irccl The code will incorporate these new features in the future. 13:29 MTDiscord Ok? 13:30 farribeiro|irccl The latest commit already supports these things. 13:31 MTDiscord Not seeing a question? 13:32 MTDiscord yes luanti will "inherit" these features on every configuration where a recent luajit version is available 13:32 MTDiscord If your asking if luanti will have these, then yes. Dependant on how its compiled, where its sourced from, etc. It's no different than goto currently, or gc64 being on or off 13:36 farribeiro|irccl Understood. I hope this will be advantageous for the project. 13:37 farribeiro|irccl It will enable the use of bitwise and other features. 13:38 MTDiscord honestly i'm not super stoked about the new features. some will be nice but out of the door it will probably be mostly a compatibility mess and modders would be better advised to stay clear of the features. 13:39 MTDiscord bitwise is already covered by bit.* and if i remember correctly, the runtime cost of that should be close to operators, so i'm not a big fan of that either. 13:39 farribeiro|irccl It will no longer be necessary to use the bit.* 13:39 MTDiscord yes, i know. i'm saying that doesn't really add anything. 13:40 [MatrxMT] it would still be needed for PUC Lua and I haven't heard any plans to drop PUC Lua 13:40 [MatrxMT] Depending on LuaJIT is a great way to make you mod not portable 13:41 MTDiscord Yes, but molders already use goto, math.random with 2? Args, etc 13:41 MTDiscord It rarely actually matters 13:41 MTDiscord i think realistically some games must depend on luajit simply because puc lua does not achieve acceptable performance 13:41 MTDiscord math.random(a, b) is standard but some edge cases around it aren't 13:41 [MatrxMT] math.random with 3 13:41 MTDiscord For pure puc to luajit, however the new features will cause more issues due to slow ass distros 13:42 MTDiscord yeah it will be a compatibility nightmare in the transition period 13:42 [MatrxMT] and then every one of those LuaJIT-only mods will eventually get user reports of the mod not working without it, often with the author not realising it would be an issue 13:42 MTDiscord also for tooling around lua. how slow will luacheck be to upgrade? VSC extensions? things like lua-lsp? selene? stylua? idk 13:42 MTDiscord Personally I care more about brain dead distros that disable gc64 13:43 [MatrxMT] maybe it could be more of a certain thing for Luanti 6.0 13:44 MTDiscord Luanti only cares about compiling luajit for non Linux platforms. On Linux it uses the system library by default so it would still get in 13:44 MTDiscord Linux, non flat pack I guess 13:44 MTDiscord honestly we should consider vendoring luajit 13:45 MTDiscord Or submoduling it..... 13:46 [MatrxMT] dropping PUC Lua seems like it's dubious before a compatibility break. A lot of downstream would expect PUC Lua compatibility 13:46 MTDiscord well yeah the specifics of how we vendor it are not that important to me, what'd be important is that luanti builds consistently use the same luajit version 13:46 MTDiscord Actual submodules, not the submodules at home shit we had with iirlichtmt 13:47 MTDiscord The puck lua is already vendored and a soft fork 13:47 MTDiscord *puc 13:48 MTDiscord But yeah, previous discussion of dropping puc lua was negative. Mainly due to the bus factor of one, and some other reasons I forget 13:48 [MatrxMT] main advantages to keeping PUC around are more cpu architectures and backwards compatibiltiy 13:49 [MatrxMT] too much asm in LuaJIT 13:51 [MatrxMT] so back to the original question: a server with a developer could use the latest version of LuaJIT and take advantage 13:51 [MatrxMT] but mod authors probably want to keep much wider/further backwards compatibility than relying on very recent features 13:52 MTDiscord Once the new luajit has been out for at least two luanti versions would be a good minimum 13:53 MTDiscord For those who dont want to be pure seedling edge/dev 13:55 MTDiscord Wonder if someone will make the equivalent of whatever it is in html/js/css land for compiling in polyfills for a certain target 13:55 MinetestBot 02[git] 04sfan5 -> 03luanti-org/minetestmapper: More cleanups and minor improvements 13ec99246 https://github.com/luanti-org/minetestmapper/commit/ec99246ef3009d42421dc5e2e1e1c250dade144d (152026-08-08T13:51:45Z) 13:56 [MatrxMT] Lua Babel with transpilation to remove gotos and turn bitwise operators into bit.* calls, ..., (not sure how it could do const) 13:58 MTDiscord const is trivial, just produce a compile error if it is violated. it's a pure load time check. 13:59 MTDiscord related: https://hisham.hm/2026/08/04/the-lua-community-needs-to-learn-to-move-on/ 13:59 MTDiscord (teal guy, teal has babelesque qualities) 14:03 [MatrxMT] right, we're talking about local scope const and not something a bit harder like class consts in Java 14:04 MTDiscord yep 14:04 MTDiscord which is also why i'm genuinely not a big fan of it as i've written on the lua mailing list but eh 14:16 MinetestBot 02[git] 04sfan5 -> 03luanti-org/minetestmapper: Add test for malformed block 13c434fe9 https://github.com/luanti-org/minetestmapper/commit/c434fe92c75a84958f62cce3d1f9e83b51737ddf (152026-08-08T14:14:05Z) 14:24 [ I prefer the openresty fork of luajit because it has ppc64le support which the upstream luajit maintainer inexplicably refuses to merge 14:25 MTDiscord ppc64le is extremely fringe, is it not? 14:26 [ it's one of the few performant options that can run with fully free firmware, plus I have one (and plan to use it as my main desktop when I get it working) 14:29 MTDiscord fringe person wants fringe thing 14:32 [ software freedom is important 14:35 [ (apart from the practical issue of ppc64le support, stuff like the luajit maintainer removing old tags and refusing to merge ppc64le and loongarch support with no public explanation whatsoever makes me distrust the luajit maintainer) 14:47 MinetestBot 02[git] 04sfan5 -> 03luanti-org/minetestmapper: Fix build-mingw.sh 1388c7381 https://github.com/luanti-org/minetestmapper/commit/88c7381a20ada1992169522d590ebad36195507e (152026-08-08T14:45:16Z) 14:52 sfan5 @wsor4035 even ignoring puc lua, relying on new luajit features can still get your in a compatibility mess with the aforementioned openresty fork, which debian uses (IIRC). no idea if they're going to pull those changes or how fast it will happen. 14:54 MTDiscord ok? 14:54 MTDiscord already knew that 14:54 sfan5 just saying 14:54 MTDiscord as luatic said, if luanti did vendor in one way or another luajit, would make things less of a mess 14:55 MTDiscord personally care about gc64 the most, since its annoying to have to deal with that issue on repeat 14:56 sfan5 there are people with disabled gc64 in the current year? 14:56 MTDiscord iirc ubuntu didnt have it enabled last one i remember 14:56 MTDiscord citation needed 14:57 MTDiscord would be happy to be wrong 15:07 farribeiro|irccl One thing I remembered, the Luanti compilation using Puc Lua doesn't work well. 15:07 farribeiro|irccl Iirc 15:11 MTDiscord "doesn't work well" isnt really descriptive 15:11 MTDiscord whats the actual problem? 15:14 MTDiscord If anyone is reading this and is confused as to the response. ppc64le is only used on IBM Power super computer servers 15:23 MTDiscord so did some research. testing on 24.04 looks like gc64 is on. it seems like this was first enabled in 22.04 or 24.04 16:19 farribeiro|irccl Wsor Several things happen, I don't know exactly what, but it's often related to Luanti's screens. 16:39 farribeiro|irccl Forget it. Using luajit luanti works 18:39 Noisytoot If anyone is reading this and is confused as to the response. ppc64le is only used on IBM Power super computer servers 18:39 Noisytoot Mostly but not quite. It's also used in some RaptorCS workstations (but IBM is still the only manufacturer of the CPUs for now, hopefully that will change (there's also some embedded ppc64 CPUs made by NXP but I'm not sure those support little-endian mode, and Microwatt that runs on an FPGA) 18:40 Noisytoot ) 19:21 MTDiscord So you're telling me that you're running a luanti server on a raptorcs workstation or a ibm power server? 19:21 MTDiscord Can I have one?