Time Nick Message 08:54 sfan5 merging #13922, #13915 in 10m 08:54 ShadowBot https://github.com/minetest/minetest/issues/13922 -- Deprecate .bmp format by sfan5 08:54 ShadowBot https://github.com/minetest/minetest/issues/13915 -- Set color of scrollbar/dropdown button symbols and checkmarks to white by erlehmann 09:32 [MTMatrix] Why #13818 is now 5.9? 09:32 ShadowBot https://github.com/minetest/minetest/issues/13818 -- Debundle Minetest Game by rollerozxa 09:32 [MTMatrix] Have we waited for nothing? 09:32 [MTMatrix] rubenwardy 09:32 ROllerozxa wtf 09:32 rubenwardy we're in feature freeze 09:33 rubenwardy not my decision, haven't been following dev lately 09:33 MTDiscord Feels worth an exception 09:33 erle hahaha 09:33 erle everything feels worth an exception 09:33 MTDiscord No 09:33 MTDiscord Just this 09:34 MTDiscord Everything else can eat dirt, but this should be merged sooner rather than later 09:34 MTDiscord It was ready to go 09:34 erle i still think this might need more research. minetest game from 5.4.1 crashes with newer minetest at load time *without any indicator that it is too old and that is the reason*, and newer minetest games might crash during runtime (i have not determined). 09:34 rubenwardy to merge #13818, you also need to merge #13906 and #13807 09:34 [MTMatrix] I 100% agree with greenxenith. That's the main UX problem MT has been carrying for years, I have to deal with that at conventions every time 09:34 ShadowBot https://github.com/minetest/minetest/issues/13818 -- Debundle Minetest Game by rollerozxa 09:34 ShadowBot https://github.com/minetest/minetest/issues/13906 -- Offer ContentDB updates for leftover bundled Minetest Game by grorp 09:35 ShadowBot https://github.com/minetest/minetest/issues/13807 -- Add package update detection on Content tab by rubenwardy 09:35 erle look the question is what feature freeze is for: is it to reduce bugs? then only nonfunctionals should be merged. is it to speed up delivery of other stuff? then it might be different. 09:36 [MTMatrix] If you're playing a many years old version, that's also kind of your fault erle. Security issues have been fixed through the years 09:36 erle saying users are at fault because they are does not make problems go away, i keep saying this. i'd at least handle crashes of minetest game specially (i.e. suggest they update) 09:36 rubenwardy feature freeze is to reduce bugs. If we merged the above we'd need to restart the freeze 09:37 erle and i think you can see from the white-interface-elements thing that something can *not* be easy (given that i failed 2 times at inverting the checkmarks correctly and did not consider them at all in the first place until prompted) 09:37 MTDiscord If it was a 5.8 milestone why was the freeze even started 09:38 erle basically everything is fizzbuzz 09:38 sfan5 we can also extend the feature freeze 09:38 erle (”we did not do it because it was easy, we did it because we thought it was easy when we started implementing”) 09:38 sfan5 or wait extra long for users to test an RC build 09:38 erle let's see if i can make a BMP slip by #13922 hehe 09:38 ShadowBot https://github.com/minetest/minetest/issues/13922 -- Deprecate .bmp format by sfan5 09:39 sfan5 or aim to release 5.9.0 quickly after 5.8.0 09:39 erle yeah 09:39 MTDiscord I am unclear on what exactly was remaining on the pr, I must have missed the prereqs on the thread 09:39 rubenwardy if we don't have this in 5.8, I'd like 5.9 for december 09:40 MTDiscord The last comments on the thread lead me to believe there was nothing blocking the pr 09:41 [MTMatrix] rubenwardy: which it's unlikely to happen, as data reveals that we are at least one-two months late for every release 09:41 [MTMatrix] *is 09:41 rubenwardy true :D 09:41 sfan5 we could also take the risk and release 5.8 as usual with the changes and perhaps be forced to quickly put out 5.8.1 09:41 rubenwardy I can update #13807 tonight 09:41 ShadowBot https://github.com/minetest/minetest/issues/13807 -- Add package update detection on Content tab by rubenwardy 09:41 erle a more-often release schedule would be nicer anyway. 09:42 MTDiscord Aha, I see the last comment was about the other pr being a suitable solution, my bad 09:42 erle btw, is there any chance to revert the keycode/scancode thing? i think there are at least 4 bugs that basically are different variants of “we detect which key is pressed now instead of which character is” 09:42 [MTMatrix] erle: I guess that'd be more stressful for devs 09:42 erle maybe 5 09:43 MTDiscord Depending on how trivial those 3 prs are though, I still believe it may be worth extending the freeze so you get the major feature everyone is waiting on sooner rather than later without having to worry about rushing a tiny release later 09:43 erle like, i can't input a +, some people can't input a / etc. pp. 09:43 rubenwardy yeah I agree greenxenith 09:43 erle it's basically a really weird regression because the thing was half-baked 09:43 erle (and if you can still input your key bindings depends on your layout basically) 09:44 erle (whereas it did work perfectly before) 09:44 [MTMatrix] Also, the lesser the UX problems, the more the feedback from non tech-savvy users who won't run away puzzled by MT/MTG duality 09:44 erle and i don't think that the fix for “+ and keypad + are considered the same key” warrants that level of breakage 09:45 erle ig what i am asking is basically, can the feature freeze be used to revert regressions if that simply gets it to an earlier state? 09:46 MTDiscord Wouldn't regressions be considered freeze-blocking bugs? 09:46 erle related: should some of the different “i can no longer input this character or key combination” issues be merged? 09:46 MTDiscord (I don't know, asking) 09:47 rubenwardy yes, regressions usually would be 09:47 rubenwardy #13890 is in the milestone 09:47 ShadowBot https://github.com/minetest/minetest/issues/13890 -- Restore the +/- numpad default behaviour (increase/decrease viewing range) 09:51 erle rubenwardy btw i think i figured out what causes the rendering corruption and why it can not be replicated. basically, at some point graphics cards reach the maximum level of instructions and they just start flushing out earlier things and everything gets corrupted. this does not ever seem to happen with mesa software rendering (that one just becomes very slow in that case), but it can, in principle, happen with every graphics 09:51 erle card. according to what i found, modern mobile GPUs will probably crap out even earlier than my old intel GPU because of the way they handle rendering. 09:51 erle you can use ”glxinfo -l” on gnu/linux to find the limits of your hardware basically 09:52 rubenwardy So, are we decided to merge those 3 PRs in the next few days, and extend the freeze? 09:52 erle so what my stubs mod does is present a worst case for item rendering (nodeboxes are pretty bad in general) 09:54 sfan5 I vote for yes 09:54 sfan5 merging #13930 too in 5-10min 09:54 ShadowBot https://github.com/minetest/minetest/issues/13930 -- Prevent Windows Defender warnings in singleplayer (Bind singleplayer server to 127.0.0.1) by rollerozxa 09:55 MTDiscord 2 core dev approval, that means it's a go, right? :)) 09:55 MTDiscord (/s) 10:04 erle has this even be unit tested 10:04 erle i mean integration tested 10:05 rubenwardy how do you plan on integration testing something that shows native windows os ui 10:06 rubenwardy rhetorical question 10:07 erle i think it will affect all other systems as well? i'll think about it though. 10:07 erle ig running a test server is not part of the pipeline 10:07 erle or is it? 10:08 rubenwardy the pipeline starts up a devtest world iirc 10:08 erle so we can be sure it works on linux at least ig 10:08 erle that's the integration test then 10:09 erle just for my understanding: binding to 127.0.0.1 will work on ipv6-only machines, but binding to ::1 would fail on ipv4-only machines? 10:12 ROllerozxa if binding to 127.0.0.1 wouldn't work, then singleplayer wouldn't have worked previously as it connects to 127.0.0.1 (see the comment on the PR from sfan) 10:13 erle ROllerozxa i see. thanks! 10:14 erle i should read more in the future ig 11:52 erle fine, i made two test cases for BMP deprecation detection myself: https://github.com/minetest/minetest/pull/13934 12:16 [MTMatrix] So can we revert MTG debundling to 5.8? 12:46 rubenwardy already done 15:59 sfan5 https://github.com/minetest/minetest/tree/ci planning to push this if CI passes 16:01 nrz_ nice 🙂 16:21 sfan5 we should probably prioritize #13913 so we can do a second string update before release 16:21 ShadowBot https://github.com/minetest/minetest/issues/13913 -- Fix issues in Minetest's English texts by srifqi 16:25 sfan5 the macOS ci apparently needs 40 minutes to install dependencies 16:25 sfan5 ah okay that's apparently due to macOS 11 16:25 sfan5 which is EOL 16:33 Desour merging #13919 in 10 16:34 ShadowBot https://github.com/minetest/minetest/issues/13919 -- [NOSQUASH] Inventory: Release resizes-locked lists on all `on_`-callbacks, and fix `Inventory *` UaF by Desour 16:45 sfan5 got it down to 1 minute \o/ 16:48 Desour did you use the powers of time, aka just wait? 20:41 erle v-rob what do you think about possibly changing back to characters? i think it would fix several regressions. i think at least desour for example mentioned having bound chat to /, which obviously also can no longer be input if “/” is not a key on the keyboard. 20:42 erle (for me it is iso level 3 shift + i, but for users of QWERTZ it is shift + 7) 20:42 [MTMatrix] I see in breakage list, you now migrate from nicknames to UUID, i have question, it can break nick based mods? 20:48 v-rob If we used characters, it wouldn't work in the SDL2 backend: it doesn't associate characters with keys (since that's taken care of by IME text events), but only gives raw keycodes and scancodes for keyboard events. 20:49 v-rob Given that we're already using keycodes, manual translation from keycode to character via a hash map wouldn't be an improvement 20:50 erle so basically minetest is in some intermediate state right now where SDL2 stuff does not fully work, but also the (previously implicit) ability to input keyboard combinations was lost 20:50 erle make no mistake, i like the SDL2 stuff 20:50 erle ctrl + shift + u and then inputting unicode is great 20:51 erle i just wonder how to handle key combinations ever again (e.g. for keyboards without a + key) 20:51 erle any idea? 20:52 erle or is the end game to do it like i3? i.e. define an initial mapping based on the key position, so it is physically the same everywhere? 20:52 erle (but then display what is on the key according to the layout obv) 20:53 v-rob Does the + keycode work in other programs? Like Ctrl + and Ctrl - for zooming in Firefox/. 20:54 v-rob I mean, at some point, it just might be a "we can't do anything about this; it's a bug for SDL to handle" 20:56 v-rob (You'd think that, by now, people would have figured out how to make keyboards standard and easy to use for developers, but nooo) 20:59 erle v-rob yes, pressing ctrl + isolevel3shift + b (i.e. ctrl + “+”) works in firefox and basically any other program. 21:00 erle v-rob what i *think* is that you can't actually bind something to modifier keys in those programs. in minetest you can bind to isolevel3shift now and that makes it impossible in principle to have it as a modifier key. 21:01 erle so what might work would exclude modifier keys (probably except shift) from having stuff bound to them 21:01 erle not sure if that is feasible 21:01 erle or even desirable 21:02 v-rob Still, SDL wouldn't generate the proper keycode, so it wouldn't solve anything. 21:02 erle v-rob if you can point me to some other SDL programs that have configurable keybindings i could try to see if it works for me with those and maybe we can steal this thing 21:02 v-rob Maybe Firefox does use characters internally? 21:03 erle yeah but look my terminal also zooms when i do ctrl + “that key combination that generates a plus character” 21:04 v-rob Gosh, I hate keyboards. 21:04 erle i think i'll just look at other games that have sdl input and handle it better than minetest 21:04 erle and then snitch on it to you ok? 21:04 v-rob sounds good 21:05 v-rob Incidentally, thanks for testing this stuff; it's good to know about these edge cases 21:06 erle well, it's not really an edge case for keyboards. for people in other countries it's just a case. i mean not for me, the neo2 layout is not german default. but there are a lot of non-QWERTY layouts. 21:07 erle when i was a child i used QWERTZ (german standard layout) and there was a lot of software that was just not handling correctly that Y and Z are exchanged compared to QWERTY 21:09 v-rob Edge cases from the perspective of " 21:09 v-rob every library that exposes keybindings for QWERTY keyboards 21:09 v-rob Which appears to be basically all of them 21:11 erle v-rob so i just tested prboom-plus which is SDL apparently and i found that for some keys that minetest misdetects, it just prints “junk” in the keybinding dialog. also i could not bind a + there. 21:11 v-rob There would be some marked benefits if everyone used the same language for everything. It's most unfortunate that programmers pretend that this is the case when it isn't. 21:11 erle i guess a lot of keys mapping to “linke taste” might just be the ”junk” response from SDL be misinterpreted? 21:12 erle (”linke taste” is german for “left button” or “left key”, depending on if you talk about a mouse or a keyboard) 21:28 erle v-rob so apparently SDL1.2 had a “unicode character” thing additionally to scancode/keycode 21:29 erle v-rob i am trying to figure out where it went 21:29 v-rob Removed in favor of IME input