Time Nick Message 02:53 MTDiscord sfan5: I looked into this.. I don't have a definitive explanation for how it is happening. But I'm very concerned. 02:54 MTDiscord It's very clear the bug is caused by a typo in SDL2Config.cmake. But the copy of SDL2 installed by vcpkg locally (at the pinned revision) has the typo fixed. But the one remotely installed has it broken. Even though both sets of config files have the same version of sdl2 02:56 MTDiscord Since this was working before, somehow the SDL2 that vcpkg is fetching has changed. Even though it should be checking hashes to verify nothing has changed 03:19 MTDiscord ok wait, the line in question is auto-generated. So maybe there's a simple explanation 04:06 MTDiscord 90% sure this is it. 4 days ago Github updated their windows runner images. They upgraded CMake from 3.29.0 to 3.29.1 04:10 MTDiscord And 3.29.1 cmake caused problems with vcpkg: https://gitlab.kitware.com/cmake/cmake/-/issues/25873 04:10 MTDiscord so the .cmake files being generated are broke 04:12 MTDiscord Here's the runner update log: https://github.com/actions/runner-images/releases/tag/win19%2F20240407.1 04:13 MTDiscord It is fixed in 3.29.2. So github just needs to update the runners again. 08:36 sfan5 @paradust the thing is: because old sdl versions did not come with a SDL2Config.cmake, our code simply imports the C header and checks the version in there 08:38 MTDiscord the code snippet doesn't compile because SDL_version.h is not found, and it's not found because SDL2_INCLUDE_DIRS is set incorrectly 08:38 sfan5 I see 08:39 MTDiscord This is a potential immediate fix: https://github.com/minetest/minetest/compare/master...paradust7:minetest:pin_cmake 08:39 MTDiscord it pins the revision to 3.28.x 08:39 MTDiscord https://github.com/paradust7/minetest/actions/runs/8657067553 08:40 MTDiscord For it to work, you also have to clear the cache entries, since vcpkg is cached, and the cache key does not include the cmake version 08:41 MTDiscord Entries are at https://github.com/minetest/minetest/actions/caches, and they begin with localGitId=... 08:41 MTDiscord i think only the latest 2 entries need clearing 08:42 sfan5 feel free to open a PR 08:44 MTDiscord I'll send a PR. I wasn't sure if it was worth it, since github will likely fix it in a day or two. 10:11 pgimeno Relevant for SDL build on Mac: https://github.com/love2d/love/commit/6a298a74c22548d1a7e259a5f9363c971362bf1f (code comment) 10:11 pgimeno In short, it appears that SDL2 is exhausting the little memory available for JIT compilation on that platform. 14:30 sfan5 merging #14536 in a few minutes 14:30 ShadowBot https://github.com/minetest/minetest/issues/14536 -- Pin cmake to 3.28 because of regression in 3.29.1 by paradust7 14:46 sfan5 hey i just found out minetest comes from a free client-side wallhack 14:46 sfan5 with a* 16:42 MTDiscord sfan5: why is minetest.get_gen_notify nil in the async mapgen environment? 16:43 sfan5 dunno 16:45 MTDiscord trying to port biomegen to the multithreaded mapgen envrionment, and local gennotify_flagstr, gennotify_decolist = minetest.get_gen_notify() Causes the following error: mapgen.lua:47: attempt to call field 'get_gen_notify' (a nil value) 16:46 MTDiscord I do see in the List of APIs exclusive to the mapgen env, a function minetest.save_gen_notify(id, data) 16:46 sfan5 dunno as in I don't remember if there was a technical reason or I just forgot 16:46 sfan5 make sure to open an issue 16:46 MTDiscord kk 16:57 MTDiscord https://github.com/minetest/minetest/issues/14540