Time Nick Message 00:27 MTDiscord Now it's -1% because even renaming types in irrlicht probably causes memory corruption by some anomalous scp level bug 01:15 proller__ no need to touh irr 04:01 MTDiscord Oh there's just so much stuff everywhere I don't even know where v3s16 comes from at this point 04:09 proller__ patch ready almost 2 years 04:20 MTDiscord what is bpos, opos? 07:44 MTDiscord @paradust blockpos objectpos 11:18 MTDiscord @panic!("is null: {}", *nullptr); Once irrlicht and minetest are in the same repo, it would be a simple matter to replace IReferenceCounted with shared_ptr (or a slightly-faster non-threadsafe version of that, if desired) 11:19 MTDiscord There are 467 calls to ->drop(). It'd be quite difficult to audit to figure out if they are being done correctly 11:19 MTDiscord And then you have stuff like this: https://github.com/minetest/minetest/blame/master/src/gui/guiFormSpecMenu.cpp#L161 11:26 MTDiscord unfortunately there might be drops that are required to happen at a specific time for side effects 13:21 proller__ paradust: look to https://github.com/minetest/minetest/pull/12142/files - to make optional break 32k limit all types should be renamed to aliases 13:24 proller__ now there is node position (s16 => pos_t) block position (s16=> bpos_t) (usually pos should be equal bpos but for better conversion control need to split ) object position (float => opos_t) 13:24 proller__ then 16/32 bit mode can be switched from one place in config 13:26 proller__ https://github.com/minetest/minetest/pull/12142/files#diff-de2d4c7a20f6da44f9046171f0be257890e7d50d78b5f2b5f3fac93cc8dd7732R68 13:46 MTDiscord That's actually evil 13:56 proller__ what do you suggest? 14:05 MTDiscord proller__, you can't see it but he was replying to paradust's "unfortunately there might be drops ... " comment 14:06 MTDiscord keep in mind that replys are not bridged 14:09 MTDiscord Oh sorry 18:09 grorp pushing https://github.com/grorp/minetest/commit/84dd812da4dc9d12282a2110a4e062848e2e8dd2 in 15 min 18:29 grorp done. 19:20 sfan5 https://0x0.st/HnGH.png looks like it might rain soon 19:22 MTDiscord Is that a teaser for a new feature or just a weird texturing bug? 19:22 sfan5 the latter 19:22 sfan5 textured clouds could be funny tho 19:41 MTDiscord Wait, can we control cloud (and star) color? Leaving alone textures, color alone would be cool. 19:44 sfan5 yes, you can 19:45 sfan5 merging #14383 in 10m 19:45 ShadowBot https://github.com/minetest/minetest/issues/14383 -- Irrlicht support changes by sfan5 19:45 Krock there's indeed some water in the sky 20:22 sfan5 I just realized that we have IrrlichtMt since 3 years ago, long time 20:48 sfan5 [47749.623437] __vm_enough_memory: pid: 59933, comm: minetest, not enough memory for the allocation 20:48 sfan5 I wonder what exactly I did to cause this 21:00 Krock malloc(-1) ? 21:06 Krock I just checked your recent irrlicht compat PR and the commit in irrlicht itself. looks good. are you sure it's a new issue? 21:06 Krock in either case - you might already be doing a bisect to figure it out 21:18 sfan5 nah it was my own freshly written code that was broken 21:44 sfan5 it seems our shader system has no concept of setting constants for specific shaders only 21:44 sfan5 so all names are shared between all shaders 21:45 sfan5 in other words if you define an uniform texelSize0 in your shader it will get a value 21:45 sfan5 if you don't, it won't but the code still tries