Time  Nick       Message
00:14 sfan5      sometimes I wonder how the awful code in irrlicht ever worked
09:45 celeron55_ a leaner release process would be nice for cases like this. now it's like "eeeh i don't want to"
09:45 celeron55_ i mean, for a 5.8.1 backport solely of the one fix
09:53 [MTMatrix] <Zughy> Hi celeron, reminder for the renaming internal discussion
10:22 [MTMatrix] <Zughy> Core devs: when you label a PR "rebase needed", please warn OP with a comment or they won't be able to know
10:22 [MTMatrix] <Zughy> at least, I don't receive any mail when a PR of mine isn't compatible anymore with master
10:35 Krock      why not use the "Pull requests" link on the top right side to check the status of all PRs instead of getting individual notifications
10:35 Krock      + ?
11:08 sfan5      if anyone is bored today please compile and test that irr#276 works on your platform
11:08 ShadowBot  https://github.com/minetest/irrlicht/issues/276 -- GL loader work by sfan5
11:45 sfan5      pushing https://github.com/minetest/minetest/commit/0f2517070e2a2e7a45ececbf88c82a717270e97f after I have acquired my pizza
12:03 sfan5      done
12:20 MTDiscord  <jordan4ibanez> Very noice on the pizza and the push
12:27 Krock      enjoy your pizza :3
12:37 Krock      compiling ....
12:39 [MTMatrix] <Zughy> Re Krock "why not use etc": there are people using GH just for Minetest (meaning the don't frequently log in), and it doesn't send an e-mail notification anyway
12:39 [MTMatrix] <Zughy> *they
13:45 sfan5      it's pretty weird how singleplayer runs the server with a target step length equal to the 1 / fps_max, which can become ridiculously small meanwhile dedicated servers run at the equivalent of 11 fps
13:46 sfan5      (i don't think singleplayer ever hits the target step)
13:47 MTDiscord  <jordan4ibanez> I suppose that's just telling the singleplayer server, run as fast as you can so this is smooth, perhaps
13:47 MTDiscord  <jordan4ibanez> But maybe it should be changed so interpolation/sync can be tested to yield better results
13:48 MTDiscord  <jordan4ibanez> Can't really be infallible data in the test bed if it's always out of sync
13:49 MTDiscord  <jordan4ibanez> Not sure if there's a server side lag simulator built into the game, that could also be neat
14:00 sfan5      I guess thinking about changing the dedicated_server_step default is also worth it
14:01 sfan5      though a basically unmodded MTG world is probably not the best test case of what real servers achieve
14:02 Krock      automation mods will surely bring down the performance quite a bit
14:03 Krock      the singleplayer step seems to be still OK. Otherwise I'd expect a 100% core usage if it were continuously looping without pause
14:05 sfan5      i was thinking of capping singleplayer step to 1/60
14:07 Krock      also physics capping to 1/60 might make sense to avoid floating point precision issues at high framerates (unless someone already fixed that)
14:08 Krock      #13400
14:08 ShadowBot  https://github.com/minetest/minetest/issues/13400 -- Mouse movement at high FPS is skewed towards the top-left corner
14:08 Krock      er no. the other one. #10962
14:08 ShadowBot  https://github.com/minetest/minetest/issues/10962 -- Client-side player movement is inaccurate at high FPS, even near the world origin
14:10 sfan5      don't think that's related
14:13 Krock      related in terms of being FPS-bound
14:13 Krock      granted, there might be better solutions to fix physics
14:26 MTDiscord  <luatic> merging #14377 in 5m
14:26 ShadowBot  https://github.com/minetest/minetest/issues/14377 -- Upgrade actions in CI by wsor4035
14:28 sfan5      do #14368 too
14:28 ShadowBot  https://github.com/minetest/minetest/issues/14368 -- Introduce std::string_view into wider use by sfan5
14:28 MTDiscord  <luatic> ok restarts timer
14:29 MTDiscord  <luatic> also lol the string view PR would basically be quadruple-approved (counting your implicit approval) if i were to add my approval now
14:31 sfan5      then I can say it's not my fault if something breaks
14:31 Krock      it can only be the compiler's fault after that :P
14:33 Krock      >  This roughly halves every 666 server steps
14:33 Krock      nice
14:34 Krock      would be even better to use std::exp to take any server step into value
14:34 Krock      (factor based on dtime)
14:35 sfan5      math was too complicated to my brain today
14:38 Krock      max_lag *= 1.0f / std::exp(dtime / 87.0f)  <-- this should reliably halve it after a minute
14:39 Krock      max_lag *= std::exp(-dtime / 87.0f)  for shorter notation
14:40 Krock      proof: exp(−10 / 87) × exp(−50 / 87) == exp(−60 / 87)
14:41 Krock      or simpler: fixed 1 second interval
17:13 rubenwardy Does GUIEditBoxWithScrollBar support colorisation?
17:14 sfan5      🤷
17:16 rubenwardy one way to find out I guess
17:17 rubenwardy it does
17:30 Desour     merging #14376 in 5 min
17:30 ShadowBot  https://github.com/minetest/minetest/issues/14376 -- Don't use a reference for RaycastState::m_pointabilities by Desour
17:48 rubenwardy Does Minetest have any utils for iterating over a utf string?
17:50 sfan5      the general solution is to utf8_to_wide(s) and iterate over that
17:50 sfan5      (still doesn't work for codepoints outside of BMP)
17:53 rubenwardy yeah that's what I'm doing currently but feels unsafe
17:53 rubenwardy in my own projects, I convert to a utf-32 string for this - I believe that covers everything except graphenes
17:53 rubenwardy updated #13825
17:53 ShadowBot  https://github.com/minetest/minetest/issues/13825 -- Add button_url[] to allow mods to open web pages by rubenwardy
17:58 MTDiscord  <luatic> rubenwardy: since you only test against ASCII characters, this should be fine
17:59 MTDiscord  <luatic> it would be fine without the utf8 -> utf16 conversion even
17:59 rubenwardy couldn't you end up inserting colors between two bytes of a codepoint in that case?
18:00 MTDiscord  <luatic> though wait yes with how it's currently written it's not fine i think
18:00 MTDiscord  <luatic> why do you have to insert one color per codepoint though?
18:00 MTDiscord  <luatic> just insert colors before and after the color-changing ascii characters. this can be done using simple regex find-replace, for instance.
18:01 MTDiscord  <luatic> this is ultimately a (very simple) tokenization problem so regex seems like the appropriate tool
18:05 rubenwardy another option would be to display the punycode https://rwdy.uk/uzNje.png
18:06 rubenwardy if the characters are all safe and ASCII, it doesn't add the xn--:     https://rwdy.uk/B3KKn.png
18:07 rubenwardy (the wikipedia domain there contains a cyrillic small a)
18:12 MTDiscord  <nrz0> legit @rubenwardy it's better to do like this (and firefox), to prevent cheating URL with unicode chars
19:14 MTDiscord  <wsor4035> sfan5: is it worth creating a (minetest/minetest) issue to track the two upstream related warnings left/vcpkg one?
19:31 sfan5      i guess
19:32 MTDiscord  <wsor4035> ok, in theory all of the minetest github org should have prs to bump versions in ci
19:59 Sokomine   recently, i'm getting a lot of connection timed out. and that not after a long time of waiting but rather soon in the process. are you already aware of this issue?
20:01 MTDiscord  <wsor4035> welcome to the past few years for windows users 🙃
20:01 MTDiscord  <luatic> merging irr#290 in 5m
20:01 ShadowBot  https://github.com/minetest/irrlicht/issues/290 -- Upgrade actions in CI by wsor4035
20:02 sfan5      Sokomine: you can likely find out the actual reason in the last few lines of a verbose client log
20:03 MTDiscord  <wsor4035> ignore me, recalled issue slightly different: https://github.com/minetest/minetest/issues/11832
20:09 rubenwardy slight problemo with the punycode solution - it requires curl 7.88+, which debian doesn't have
20:10 MTDiscord  <luatic> is there a good C++ punycode library we could vendor
20:11 sfan5      implementing punycode is probably 50 loc
20:11 sfan5      but why all this anyway?
20:12 rubenwardy it's to allow a user to preview a URL before opening it, without allow homograph attacks
20:12 rubenwardy https://en.wikipedia.org/wiki/IDN_homograph_attack
20:12 rubenwardy #13825
20:12 ShadowBot  https://github.com/minetest/minetest/issues/13825 -- Add button_url[] and hypertext element to allow mods to open web pages by rubenwardy
20:12 sfan5      can't we move the responsibility to the browser? obv not as good but browsers have this figured out and will display the url safely in the top bar
20:13 rubenwardy when it's open in the browser it's probably too late
20:15 sfan5      well not really
20:22 rubenwardy The reason to show the URL to the user is so they can decide whether to open it or not, if a dodgy URL is already open then there's a risk of installing malware or being phished if the user isn't observant
20:25 rubenwardy but perhaps if curl is too outdated or unavailable, we could get away with adding a warning. All our official builds have recent curl versions
20:27 sfan5      different question: what prevents the url from simply redirecting the user?
20:28 rubenwardy nothing, but the user can see if it's see if it's wikipediа.org or wikipedia.org
20:28 sfan5      i guess
20:29 sfan5      how about escaping unicode characters if IDN support isn't available? like wikipedi<0B12>.org
20:29 sfan5      looks strange but most people don't use unicode in domains anyway
20:45 rubenwardy https://rwdy.uk/Mt1eI.png
20:46 sfan5      looks good
20:49 rubenwardy Do you think I should use punycode if support is available, or always just render like this. The latter would result in simpler code and unit tests
20:49 rubenwardy simpler is probably better I guess
20:50 rubenwardy although goes back to always parsing unicode
20:50 sfan5      urlencode just takes the utf-8 bytes
20:50 sfan5      or what do you mean?
20:51 rubenwardy I'm iterating over the hostname to generate this colorize text
20:52 sfan5      you can treat it as a byte string in this case
20:52 sfan5      everything that's not ascii goes through percent encoding
21:37 rubenwardy lol, the textarea's colorize support is pretty broken, it shows colorize code if you try and highlight text
21:56 MTDiscord  <nrz0> @sfan5 it's more the same thing than discord client, and most chat apps tend to do nowadays, it show the domain name. Maybe we should just show the domain name @rubenwardy ?
22:03 rubenwardy that doesn't solve any thing?
22:03 rubenwardy like, it doesn't solve the problem we were discussing - with homographs
22:04 rubenwardy but maybe you don't want to overload the user with too much info
22:05 rubenwardy here's what it looks like on Discord https://rwdy.uk/jZq3s.png
22:14 MTDiscord  <luatic> Discord punycode-escapes domain names though, it seems.
22:14 MTDiscord  <luatic> https://cdn.discordapp.com/attachments/747163566800633906/1208536964295430234/Screenshot_from_2024-02-17_23-13-25.png?ex=65e3a4c1&is=65d12fc1&hm=8412f7a9048a2c2587abd7e91c6299efc2f33c9967e99218aabd5fd54834ab52&
22:25 MTDiscord  <mnh48> it seemed like Discord just punycode-escape the domain name itself, while rendering the later part of URL as unicode instead of percent encode
22:26 MTDiscord  <luatic> Yes, only the domain name is really relevant, because that's where the user's trust (ought) to come from.