Time Nick Message 08:47 Krock I don't think non-SDL has seen any notable changes since 5.8.0 08:47 Krock irr device win32/osx and linux weren't changed since the import 08:54 MTDiscord What about other code, though? What code may depend on SDL? 08:54 sfan5 nothing else 08:54 Krock there's no macros 08:58 SFENCE Btw, I was able to make minetest.app builded for arm64 working by command xattr -cr minetest.app 08:59 SFENCE it is the workaround for unsigned application. 09:00 SFENCE if we add that info to web, we can distribute also mintest.app arm64 build. 09:00 SFENCE maybe add info for intel build, that it have to be added to "trusted" application can be also usefull. 11:44 SFENCE #14947 ready for review. If we wants Mac ARM app, it can be added to 5.9.0 milestone. 11:44 ShadowBot https://github.com/minetest/minetest/issues/14947 -- Try to build universal minetest.app and arm app. by sfence 16:35 [MTMatrix] it'd be nice if some core dev called the shots and we reached a decision https://github.com/minetest/minetest/pull/14944#issuecomment-2280273086 16:39 [MTMatrix] sfan5 what would you do? 16:50 celeron55 is an official vote needed or should we just get a round of comments onto that issue? 16:51 celeron55 (i mean, PR) 16:51 celeron55 should we agree on what the primary goal is now? i propose the primary goal is to get the release out ASAP, and my opinion is that 14944 is the way to do that 16:52 celeron55 (i'll post this on the PR also) 17:11 rubenwardy if we decide to ship with SDL2, we should at least change the binding of + to work on QWERTY keyboards 17:17 rubenwardy But please, whatever we do can we release in the next 7 days 17:21 celeron55 i don't like the idea of somebody essentially getting blamed for breaking keyboard bindings. they can be broken once it's possible to list some pros in addition to the cons 17:25 sfan5 + and - seems to be the biggest issue, we could add a workaround for these two in particular 17:26 sfan5 releasing without sdl2 doesn't look attractive to me 17:29 Krock funnily enough, the forum poll doesn't bring us any further 17:30 Krock (mine's included there) 17:33 Krock Either way, I will continue to use Minetest without SDL2. Less dependencies and as a bonus, the hotkeys work as-is. 17:36 celeron55 sfan5: it's clear what's not attractive about SDL2 and what's attractive about irrlicht, but to you, what exactly is attractive about SDL2? 17:37 rubenwardy Could you bind to = instead of +? 17:38 rubenwardy I agree that it would be a shame to release 5.9 without sdl2 but it's also a shame to be in feature freeze for 2 months 17:41 Krock = is Shift+0 on my end, thus won't work with USE_SDL2=1 17:42 [MTMatrix] period, comma? 17:43 celeron55 releasing without sdl2 is not about giving up, it's about getting the release out of the way and bringing back the normal development atmosphere where things are best solved 17:46 Krock I am under the impression that we're justifying features as a trade-off for a bug in gameplay. 17:46 MTDiscord I thought the forum poll was 'give up on SDL forever'. It makes sense to disable it for release builds if it could cause bad user experience. 17:46 Krock the SDL2 code will not vanish. In contrary. 5.10.0-dev should be used to clean up the remaining bugs and then switch to it permanently to purge some old code. 17:47 Krock .. in long term. 17:50 MTDiscord That sounds better, especially if it turns out there are a bunch of other bugs with the 5.9 release 18:13 MTDiscord As a non core dev who holds no sway: 1. Yes we'll be moving to SDL2 long term, that seems valuable. 2. That is on no required timeline. 3. We should pull it out of 5.9, take some blame and move on to hammering out the buggy details of key handling. 4. Nothing for 5.10+++ is set in stone 18:15 [MTMatrix] So #14944, rc2 and let's release in a few days so it's over? 18:15 ShadowBot https://github.com/minetest/minetest/issues/14944 -- Disable SDL2 for 5.9.0 by rubenwardy 18:30 v-rob Who is at fault for all this nonsense?! I can't figure out the reason for this from git blame: https://forum.minetest.net/viewtopic.php?p=438234#p438234 18:31 MTDiscord is this even with padding[]? 18:33 [MTMatrix] fixed size formspecs don't care about padding[] I think? 18:33 Krock is it literally 800x600 or just something close to that? 18:33 [MTMatrix] v-rob: remove it! 18:34 v-rob It's literally 800x600. padding[] has no effect on fixed size formspecs 18:34 v-rob I hat formspecs so much. So much. 18:34 v-rob *hate 18:34 Krock > lockSize(true,v2u32(800,600)); 18:35 v-rob People went out of their way to add this functionality. Why? Why?? 18:35 [MTMatrix] formspec scaling ... btw, #14840 18:35 ShadowBot https://github.com/minetest/minetest/issues/14840 -- Improve formspec scaling on small screens by grorp 18:35 Krock this code is 10 years old 19:13 rubenwardy v-rob, Krock: yeah I discovered this when doing max_formspec_size, which is why it's for non-fixed 19:16 v-rob It's seeing things like this that keep me motivated to finish the new UI 19:22 sfan5 celeron55: nothing about concrete features just doesn't feel good to roll it back on short notice 19:24 celeron55 there is the question: when will 14944 be reverted in the master branch 19:25 celeron55 right after the feature freeze? or along some future PR that relates to keybindings? 19:25 celeron55 or right before the next feature freeze 19:26 sfan5 right after the release would be best 19:27 celeron55 rubenwardy should probably define it in the PR description 19:27 SFENCE If we are targeting to use SDL2 in 0.5.10 release, we should do it imidietly after release. We will have more time to test if everything works well with SDL2. 19:50 [MTMatrix] so 2 core devs in favour (sfence, v-rob) https://github.com/minetest/minetest/pull/14944, I guess ruben as well since he's opened the PR in the first place + Krock here makes 4 19:51 [MTMatrix] celeron55 too actually. So, let's go..? 19:52 [MTMatrix] RE v-rob: "[...]keep me motivated to finish the new UI" 19:52 [MTMatrix] Please do 🙃 21:15 [MTMatrix] Desour and grorp have left a few +1 in that PR, I'd say it's settled then 22:04 MTDiscord I support it as well. The keybinding issue hasn't been the only bug related to the SDL driver (though other bugs might be easier to address), and won't be the last bug either. Let's not rush SDL for 5.9. 22:14 Noisytoot what's 0.5.10? do you mean 5.10? 22:18 Noisytoot (repeating since SFENCE was offline when I asked) SFENCE: what's 0.5.10? do you mean 5.10? 22:28 rubenwardy any idea how to test whether Minetest is doing http/2 properly, reusing connections 22:28 rubenwardy could use a httpserver/logger that reports requests and connection open/close 22:59 sfan5 your access log + tcpdump 22:59 sfan5 same port = connection reuse and at least with nginx I can tell from my log if a request has h2 22:59 sfan5 was* 23:18 rubenwardy Ok, Minetest does do HTTP/2 properly 23:18 rubenwardy and does default to /2 on newer libcurl 23:23 rubenwardy I do see most requests using http/1.1 and only a couple on http/2 23:23 rubenwardy HTTP/1.1" 200 "-" "Minetest/5.8.0 (Windows/10 23:24 MTDiscord curl doesn't get built with libnghttp2 on windows and android 23:25 MTDiscord (but when you use your system's curl library on linux it will usually support http/2) 23:25 rubenwardy ahh 23:25 rubenwardy it'll probably make a notable different to lots of remote media 23:26 rubenwardy loading up to 12 CDB screenshots maybe not as much