Time Nick Message 06:19 MTDiscord Does anybody know why the standard Irrlicht materials like EMT_SOLID, EMT_TRANSPARENT_ALPHA_CHANNEL and etc in the OpenGL driver don't use own shaders when their materials renderers are being added, but only configure OpenGL state? It doesn't make possible to set and send additional vertex attributes if use derivatives of S3DVertex, e.g. through glTexCoordPointer as a hack 06:21 MTDiscord And would it be acceptable to fix it by compiling shaders for them like in the OpenGL3 instead? Like what now my two-color vertex format PR makes 09:37 sfan5 00:23 forget snooping, mitm is the issue 09:37 sfan5 tls is a thing and we're using it 09:38 sfan5 Oh it does that's wonderful, if it's already a dependency why is client traffic not encrypted? 09:39 sfan5 even if we were directly interfacing with SSL libs (which we're not), it's more than just slapping e.g. DTLS on it 09:40 sfan5 not to mention that this doesn't prevent MITM unless you also make certificates mandatory somehow 09:40 sfan5 many Minetest servers don't have a domain so you'd end up with something like TOFU (trust on first use) 09:41 sfan5 however then you buy yourself problems of what if the server rotates its key and the UI to communicate this to users in a way that they understand it 09:42 celeron55_ encryption is easy(ish) as it's just technology, identifying the other party is hard because you need to have a mandatory process which can involve an authority of some kind, or users not being dumb, or both 09:43 celeron55_ web browsers solve this by basically not actually identifying anything to a useful extent. anyone can get a certificate and play a trusted party 09:45 sfan5 @andrey2470t that is simply because the opengl driver is built on the legacy fixed pipeline. fixing this is a dead end, so we will have to drop support for the opengl driver if we want to do this 09:45 celeron55_ the thing web browsers use mostly just makes sure requests from a website go to only servers the website owner wants to allow them to go to 09:45 celeron55_ even in a mitm situation 09:46 celeron55_ except that not really 09:46 sfan5 dropping support for opengl without shaders has been brought up a few times and is basically a done deal 09:53 rubenwardy The server list could pin/preload certificates for public servers and then TOFU for private servers 09:54 rubenwardy although not sure how you'd verify from the server to the server list 09:54 sfan5 https 09:54 sfan5 ah no 09:54 rubenwardy server list needs to know the sender actually the server 09:55 rubenwardy using domain certs like letsencrypt is a nice solution for servers that have domains 09:55 rubenwardy there's also SRP - it allows you to verify the server is the same one that you signed up on 09:56 rubenwardy well, or the db has been compromised 09:56 celeron55_ well, the server list kind of is the dns of minetest servers 09:57 rubenwardy SRP is a key exchange algorithm, not sure on its suitability of setting up encrypted traffic 09:57 rubenwardy +But 10:31 MTDiscord Is the dropping of the opengl driver the ongoing process now? Then maybe is it not necessary to support this driver in my PR at all? 12:42 sfan5 it's not ongoing but cold be done at any time 20:18 pgimeno about #14830, what is the kernel version useful for? 20:18 ShadowBot https://github.com/minetest/minetest/issues/14830 -- Minimize data sent in HTTP user agent 20:19 sfan5 rough, general indicator how recent the user's OS is I'd say 20:21 sfan5 (doesn't work on android where vendors use a terrible mix of outdated stuff) 21:54 Noisytoot Why is the value returned by core.camera:get_look_horizontal() 90 degrees offset from the yaw in debug info, and the value returned by core.camera:get_look_vertical() inverted from the pitch in debug info? They're also both in radians while debug info and the values passed to Game::updatePlayerControl are in degrees, and the pitch passed to Game::updatePlayerControl is also inverted from debug info but the yaw is the same. 22:38 MTDiscord sounds like normal MT: as in the implementer wanted it that way 22:51 fishmongler hi, do we have any CI builds from 2022 or the ability to recreate them? 22:51 fishmongler for #14818 22:51 ShadowBot https://github.com/minetest/minetest/issues/14818 -- Attachments lag behind bones under some circumstances 22:54 MTDiscord someone with the right privs on github could rerun ci for a commit 22:54 fishmongler trying to build some of those commits on debian but finding the correct irrmt version for each is a time sink 22:54 MTDiscord is there a reason not to use a build from releases? 22:54 fishmongler and ideally i'd like a windows build (and i don't want to recreate that mingw setup yet again) 22:54 fishmongler i need to bisect something inbetween releases 22:55 fishmongler some commit broke attachments, maybe in minetest maybe in irrmt 22:55 MTDiscord ah, then you would need to ask someone to rerun ci for the given commit, or else just fork off at a certain commit and push to github to run ci 22:57 fishmongler in the meantime i'm torturing 5.5 and there's definitely no trace of the bug there 22:58 fishmongler problem is not a single commit in that range looks like it touches relevant code