Time Nick Message 01:48 MTDiscord Re: the oft-rumored "entity rendering is slow" problem ... is there an existing issue for it in GH already? I looked for one but only found one for particles, not entities. 01:49 MTDiscord It has been suggested to me that this may be because people assumed that you'd run into severe server-side performance problems with many entities long before client-side became a problem, but I've actually already encountered (and worked around) this problem in the real world, so I'd like to have an issue open for a possible proper fix someday. 01:50 MTDiscord If I DO end up having to open a new issue for this, is it a bug (like "entity rendering slow") or a feature request (like "make entity rendering more optimal")? 03:00 srifqi hello. #13872 is ready for review. the APKs are not ready yet 03:00 ShadowBot https://github.com/minetest/minetest/issues/13872 -- Formspec: Pass the second-touch event as is by srifqi 03:24 srifqi ah, the VS checks are failed due to vcpkg 03:40 srifqi the APKs are ready for testing (or use the one from GitHub action) 04:52 erle warr1024 i would be very amused (and slightly irritated) if someone would tell you that opening an issue about a long-standing performance issue would be a feature request. or was this sarcasm, given that ”more optimal” is … a bit weird? 11:43 Desour warr1024: I don't think we have an issue for that. (issues about the quality/performance of certain parts of the engine are generally rare. probably because it feels so in-concrete) if you can't find one either, please open it. I don't care if you take the bug report or feature request template, bug report might be more suitable for reproduction steps 11:58 MTDiscord Ultimately "performance issue" is a third category; it's neither "bug" nor "feature request". 12:03 [MTMatrix] minetest 5.8.0 in 2024? 12:53 [MTMatrix] localhost: a couple weeks I guess 13:36 rubenwardy 5.8.0 in November 13:37 rubenwardy (if we release in november, it can be used in the game jam) 14:46 [MTMatrix] I host a workshop in December, it'd be great to have people interact with MT 5.8 14:48 erle oh, what kind of workshop and where? is it online? 14:58 grorp Hi. Regarding #14007: I have authored 2 Irrlicht PRs and 2 Minetest PRs about texture filtering since the 5.7.0 release. These PRs have resulted in bugs and are causing a lot of unnecessary drama on IRC and Github. 14:58 ShadowBot https://github.com/minetest/minetest/issues/14007 -- Texture filtering regressions (biilinear and trilinear filtering broken, anisotropic filtering results inconsistent depending on GPU) 14:58 grorp I don't currently have the time or the motivation to investigate and fix these bugs. For this reason, my approach would be to revert the relevant PR(s). 14:58 grorp If someone else wants to investigate and fix these bugs, a revert is of course unncessary. 14:59 rubenwardy I haven't had time to look into these bug reports, but if confirmed then I'm fine with a revert 14:59 erle grorp what is the second PR? 14:59 rubenwardy (with a caveat of the minetest filters being in MT) 15:01 erle like we have https://github.com/minetest/minetest/pull/13683 is the second one this change? https://github.com/minetest/minetest/commit/9bef3c136a12c9238d6a55b5c8e14fc9b779db61 15:02 erle Warr1024 would you please check if the code in https://github.com/minetest/minetest/pull/14006 fixes your screenshot isuses or not? i feel unqualified to determine if it still occurs. 15:09 sfan5 for this release it should definitely be reverted 15:09 sfan5 I think the idea to get rid of bilinear for upscaling was good however 15:36 MTDiscord I am also in favor of deprecating bilinear upscaling, as long as we don't trash downscaling in the process. If we eventually ever achieve a modernized shader-driven rendering pipeline (assuming it doesn't bar me from running MT altogether) then there are actually good "crisp" upscaling algos that preserve the squareness of pixels but offer intra-texture antialiasing without blurring that would be really nice to see. I've seen at 15:36 MTDiscord least webgl demos of these things so I know they're achievable in principle. 15:37 MTDiscord In the meantime, I can definitely live with upscaling without antialiasing; the "jagged edges" of pixels within a texture are at least consistent with the edges of polygonal geometry without AA (and we have other AA options for people who have the horsepower for it). 15:37 erle refering the “squareness” of pixels is something that implies that the art stdle where pixels are point samples that need interpolation is not welcome (a POV which i can understand, i just want to point it out) 15:38 MTDiscord Linear upscaling as an artistic choice option would be nice, but it would really be an enhancement because we've never actually offered creators control over this... 15:38 MTDiscord and tbh linear upscaling is a shitty implementation of "pixels as point samples"; there are much better models, like gaussian, cubic, or sync, which better capture it 😄 15:39 erle which i think would *also* not be desired 15:39 erle if we go with the art style justification 15:39 erle btw i think the question if filtering is desired by users can probably be figured out by looking at screenshots and comparing it to other non-default options that are turned on in screenshots. 15:40 erle (most users go with defaults, same as most devs go with defaults, so the question for me here is: is this enabled at the same rate as other non-default graphics options are? maybe more? maybe less?) 15:41 MTDiscord oh, ick, we have to revert both MT and IrrMT in sync? That makes it majorly more complicated for me to compile 😒 ... it would have been so much nicer to just migrate the IrrMT stuff into MT, even if it does make the change inherently riskier... 15:41 erle the entire irrlichtmt situation is homegrown. but if you can tell me WHERE to place this with zero risk of regressions, please do. 15:42 erle my POV right now is: “i have no more experience than others which spent more time on coming up with it, any change i do that is not exactly like it was before has a non-zero chance of breaking something” 15:43 erle especially because this stuff was approved and has no testing, i would not have much faith in an approval. but tell me, where would you put that filter function logically? 15:45 MTDiscord I don't know the solutions for any of this stuff, I'm just pointing out that obstacles to testing are a problem. If somebody who knows the system better knows a solution, then all other things being equal, I would like it. If somebody does not, then I guess I'll take the L, it'll just be a bit longer before I figure out how to test certain things. 18:01 sfan5 Announcement: 18:01 sfan5 Per an internal team discussion and vote (6Y 1N) I am announcing that erlehmann will be banned from participating in minetest-dev IRC and the Minetest Github Org for one year, effective immediately. 18:21 [MTMatrix] Reasons: 18:21 [MTMatrix] 1. forcing development to spend a disproportinate amount of time on minor issues 18:22 [MTMatrix] 2. lots of talking, very little dev work actually done 18:22 [MTMatrix] 3. walls of text on IRC with the tendency to derail conversation 18:22 [MTMatrix] 4. passive-aggression reported by multiple staff members 18:22 [MTMatrix] All these aspects had been reported to erlehmann in the past, multiple times; unfortunately sorting little to no effect 18:22 nrz Thanks for the info. He talk so much that it confuse people 18:22 [MTMatrix] *disproportionate 19:04 sfan5 can I push https://github.com/minetest/minetest/commit/0e4de28988a96e87f6391a654eff8eaeed3ef801? 19:05 celeron55 seems harmless 19:05 sfan5 done 19:11 MTDiscord In the dreaded #14007, it looks like we (at least I) may have gotten 2 different issues mixed together: there's the loss of linear filtered upscaling (a controversial "feature" that was intentionally removed) and the loss of filtered downscaling on at least some hardware (seemingly an unintentional consequence). Should these be split into separate issues? 19:11 ShadowBot https://github.com/minetest/minetest/issues/14007 -- Texture filtering regressions (biilinear and trilinear filtering broken, anisotropic filtering results inconsistent depending on GPU) 19:12 sfan5 they should 19:12 MTDiscord I think it was my mistake; I misintepreted the issue when I first read it, but it does look like it was originally specifically about the upscaling. 19:13 MTDiscord I can open a new issue, transfer my screenshots and whatnot, and then I'll just mark my comments on the original as off-topic (I think I have the ability to modify my own, right?) 19:16 MTDiscord unless we're just gonna do the revert for 5.8 already (making it moot), in which case I suppose I should just wait and see if it gets reintroduced in 5.9-dev. 19:16 MTDiscord Let me know which you prefer. 19:17 sfan5 it will be reverted before release 19:19 MTDiscord Thanks! I'll just leave it alone then. 19:34 Desour 14007 is also about a 3rd issue: anisotripic filter causes bilinear filtering on some platforms, for which the texture size min is a workaround 19:34 MTDiscord Oof, I didn't even notice that was a separable issue. That sucks. 19:37 sfan5 merging #14009, #14002, #14000, #13974 in 5m 19:37 ShadowBot https://github.com/minetest/minetest/issues/14009 -- [NOSQUASH :3] Fix some invalid float to integer conversions by Desour 19:37 ShadowBot https://github.com/minetest/minetest/issues/14002 -- Fix openSUSE build dependencies by jordan4ibanez 19:37 ShadowBot https://github.com/minetest/minetest/issues/14000 -- Add tick marks to position HUD element docs by MisterE123 19:37 ShadowBot https://github.com/minetest/minetest/issues/13974 -- Fix mod translation updater bugs by Wuzzy2