Time Nick Message 11:30 sfan5 src/client/mapblock_mesh.h:267: typedef std::pair MeshIndex; 11:30 sfan5 Unsupported combination of multiline comments. New multiline comment should begin on new line. 11:30 sfan5 can someone explain this? what's the problem here? 11:47 pgimeno what gcc version? 11:49 sfan5 not gcc, this check on CI https://github.com/luanti-org/luanti/blob/081624a229b8f98f624df7e410467a528f1d7b74/.github/workflows/whitespace_checks.yml#L53 11:49 sfan5 i see the reason is written there 11:49 sfan5 but I don't agree 11:52 pgimeno unsupported by whom, then? 11:52 sfan5 util/ci/indent_tab_preprocess.py or the regex check after that, I assume 12:23 Krock that restriction makes no sense 12:24 Krock or rather - it's not made for this use-case. 17:17 sfan5 are there known issues with gui scaling/hidpi on gnome? 17:18 sfan5 when Luanti launches it is tiny on this laptop 17:18 sfan5 (missing the 2x scale factor, probably) 17:21 sfan5 with display_density_factor = 2 it looks normal, but also that doesn't affect the default window size 17:23 Krock the display size is something we define outselves, thus need to add code there 17:23 Krock s/display/window/ 17:23 sfan5 https://github.com/luanti-org/luanti/blob/081624a229b8f98f624df7e410467a528f1d7b74/irr/src/CIrrDeviceSDL.cpp#L647-L656 i think this should handle it actually 17:23 Krock I don't have a hiDPI display, thus idk. 17:23 sfan5 but only if SDL were correctly detecting the scale 17:23 Krock how about the irrlicht devices? 17:24 Krock not handled, from what it seems. 17:26 Krock also why are we dividing by `ScaleX`? Multiplying would make more sense. 17:29 sfan5 the X11 device isn't any better 17:29 sfan5 this might be yet another of those sdl2-via-sdl3 bugs 17:35 Krock I have an SDL3 PR if you'd like to test that too :3 17:39 Krock I've got a design question here: https://github.com/luanti-org/luanti/pull/16538#discussion_r2443397321 - I'd welcome inputs. 17:39 Krock ^ "Should we bump already, or in the 5.15.0-rc build?" 17:54 Krock Another unrelated question: how come the libsdl[23]-dev Debian/Ubuntu packages depends on all headers required to build SDL programs? Its documentation states that required libraries are loaded during runtime, and ldd confirms that. 17:54 Krock s/SDL programs/SDL itself/ 17:55 Krock hence, the headers of libraries detected during runtime should not be needed at all 18:03 [MatrxMT] I think it's for an SDL(2?) header file that also includes platform-specific headers 18:37 Krock 27 direct dependencies. hmm. 18:43 sfan5 does anyone here know why we have MATERIAL_FLAG_CRACK_OVERLAY? 18:44 sfan5 it changes a tiny detail in how crack pixels are overlaid onto the base texture and is apparently used by fencelike 18:46 [MatrxMT] iirc, sdl2 only does detects scaling factor correctly on wayland, not x11. iirc, they extended that to x11 in sdl3 (unsure). idk about sdl2-compat 18:47 [MatrxMT] s/sdl2 only does/sdl2 only/ 18:47 sfan5 this is wayland 18:48 sfan5 anyway I might investigate this some other day 18:50 [MatrxMT] btw, regarding y5nw's ping on the SDL3 PR: I did some testing with pre-stable-release sdl3, but I don't think there's anything salvageable from that 19:06 sfan5 Krock: is that why you were concerned about SDL dependencies? 19:08 cx384 sfan5: It looks like MATERIAL_FLAG_CRACK_OVERLAY is used to for transparency it activates this https://github.com/luanti-org/luanti/blob/081624a229b8f98f624df7e410467a528f1d7b74/src/client/imagesource.cpp#L820 19:10 sfan5 i checked again and MATERIAL_FLAG_CRACK_OVERLAY is actually the default for useTile() 19:10 sfan5 so it should be used everywhere 19:11 Krock sfan5: not just that. It is also not yet very clear to me what the mandatory compile-time dependencies are. 19:13 sfan5 I can tell you that on windows it's zero deps 19:13 Krock well yes. The Windows API provides everything through dinput/dsound and so on 19:14 sfan5 (at least the one I build for luanti. it has audio disabled) 19:14 sfan5 I would assume linux to be similar provided you build from source. distro packages typically enable everything 19:15 sfan5 https://github.com/sfan5/mingw-pkgs/blob/master/sdl2 for reference 19:15 Krock SDL is customized by generic SDL_* CMake variables, such as "SDL_OPENGL". However, only few of them are actually needed by Luanti. However, that is yet not documented. 19:15 Krock and I couldn't find any SDL documentation on the exact dependencies implied by those variables 19:16 Krock it surely is convenient if distros compile it with everything, but I can only hope that'll always be the case - for all features that we'd make use of. 19:17 sfan5 that's why they do it :) 19:17 sfan5 cx384: so I checked what happens with MATERIAL_FLAG_CRACK_OVERLAY forcibly enabled (it apparently actually isn't), that breaks crack on semi-transparent nodes 19:18 sfan5 so we don't want that 19:31 MTDiscord "how come the libsdl[23]-dev Debian/Ubuntu packages depends on all headers required to build SDL programs" Krock: debian's SDL packaging disables SDL's dynamic loading of libraries and turn them into explicit linkage, I assume it has something to do with that. a proper SDL library built from upstream on linux does not explicitly link to anything other than libc and ld if you run a ldd on it (it will then dlopen anything else if it's 19:31 MTDiscord available on the system) 19:33 Krock @rollerozxa I see. My locally compiled libsdl3 does also only have few dynamic linker dependencies. So that non-default configuration might explain it. 19:33 MTDiscord yeah 19:34 Krock linux-vdso.so libm.so libc.so and ld-linux-x86-64.so to be exact