Time Nick Message 00:34 MTDiscord AI, without actual code context suggests it could be in our texture coord calcs, not raw geometry, which I mean, sounds like an off by one error somewhere 00:35 MTDiscord See if it can fix it, that's the real test. Half the time I try to fix something that looks like an off-by-one error, I find out it's off by a bigger one in the other direction after. 00:35 MTDiscord Will try it later tonight, if my family can get me a solid hour of time, ha 00:54 MTDiscord Ha, on second ask, after tossing our inventory shader into it, pointing me to our world projection matrix, saying we might be using perspective instead of orthogonal, that also sounds promising 01:42 MTDiscord Okay, thank goodness, can replicate on windows with current master and changing opengl2 to opengl 3 does cause it to go away and come back - which means I can test fixes actualy 02:11 MTDiscord Damn, I think that actually worked and the solution makes sense. Will post to the issue 02:27 MTDiscord https://github.com/luanti-org/luanti/issues/15524#issuecomment-3530548583 Alrighty, someone tell me it's wrong lol 09:15 MTDiscord whoa thanks! a bit surprised as to why it's .375 though, where does this come from? I would've expected .5 or .25 09:58 MTDiscord Well, tbh could try out a few offset values. 09:58 MTDiscord It is likely that anything in some range works 11:43 MTDiscord I went with what we already used, so it's identical behavior instead of maybe better but different 11:45 sfan5 merging #16670, #16675 in 15m 11:45 ShadowBot https://github.com/luanti-org/luanti/issues/16670 -- [no squash] Clean up some GLSL code + more CI by sfan5 11:45 ShadowBot https://github.com/luanti-org/luanti/issues/16675 -- Remove EventReceiver by cx384 12:10 sfan5 done 17:28 sfan5 don't trust AI answers, exhibit A: https://0x0.st/KpyG.png 17:36 [MatrxMT] Interesting topic you have there 👀 :) 17:39 Krock We had one thing called specification 17:39 sfan5 yeah it's 363 pages (excluding GLSL) 17:40 Krock I'm glad that's more compact than the x86 specification 17:45 MTDiscord Curiously, AI erroneously stated to me that the SO_CLOSOCK option is not part of the POSIX standard, and it was actually quite difficult to get the AI to admit that it is, even asking it to reference the standard. It makes me wonder if these AIs somehow haven't had official standards for things included in their dataset. 17:45 MTDiscord Sorr, SO_CLOFORK. 17:45 lhofhansl Merging #16671 in a few 17:45 ShadowBot https://github.com/luanti-org/luanti/issues/16671 -- Force highp precision for float and sampler2d on GLES, if supported by lhofhansl 17:46 sfan5 > POSIX 17:46 sfan5 > POSIX.1-2024 specifies O_CLOFORK, but Linux doesn't support it. 17:46 sfan5 huh 17:47 MTDiscord It was also shocking to me that Linux doesn't support it, but that's a different rabbit hole lol. 17:49 MTDiscord I think I was talking to Copilot in the Edge browser. Some other AIs might do better. 17:53 lhofhansl Done... And off to work. 17:55 MTDiscord Base copilot in edge is garbage lol. Like worse than what I can run at home on my 1660. That said, still used it for syntax and quick lookups, so it's not bad as a first page. But yeah: all AI usage must be testable, ideally before it's shipped haha 18:05 sfan5 I used the web UI, not sure if that's different 18:06 sfan5 (copilot.microsoft.com) 18:16 MTDiscord Nope. If you don't hate Google, best free service (with a Google account) is at aistudio.google.com But otherwise, I find Mistral to be pretty decent also on their free tier 18:18 user333_ all AI must die! 18:18 user333_ otherwise we continue to get trash PRs like https://github.com/luanti-org/luanti/pull/16447 18:29 MTDiscord Ai is a tool, it depends on the user tbh 18:37 user333_ it should not be used for programming at this rate though 18:37 user333_ (among other things) 19:11 sfan5 @grorp do you want to do the "framebuffer suitability check" in a followup PR or the already open one? 19:23 MTDiscord @grorp 19:49 [MatrxMT] [@grorp:matrix.org](https://matrix.to/#/@grorp:matrix.org) ;) 19:51 Krock I though it would suffice for Matrix to just mention their name without a prefix 20:14 [MatrxMT] Generally yes. If you write Zughy I get pinged