Version bump for minetest. Ebuild and patches attached. There was some debate over whether unchanged patch files should be bumped with each version in bug#534008. That issue is moot here because the patches in question required changes in this version. Parts of each patch were adopted upstream. Reproducible: Always
Created attachment 410486 [details] Ebuild for 0.4.13
Created attachment 410488 [details, diff] Updated as-needed patch -- smaller now since some changes were adopted upstream
Created attachment 410490 [details, diff] Updated shared-irrlicht patch -- smaller now since some changes were adopted upstream
Whoops! title should be for games-action/minetest, not minetest_game. minetest_game just needs a regular version bump.
I think there are some upstream changes to the cmake flags. DISABLE_LUAJIT was removed and we should use ENABLE_LUAJIT to control this. There are also new flags ENABLE_SYSTEM_GMP and ENABLE_SPATIAL which should be added to use flags. There is also a small problem that is also in 0.4.10 ebuilds. We should pass -DCUSTOM_EXAMPLE_CONF_DIR="/usr/share/doc/${PF}" to cmake or it will install a single file into /usr/share/doc/minetest.
Created attachment 410546 [details] New ebuild with added flags Thanks for the suggestions. This ebuild makes those changes as well as adding the gles flag. I didn't actually test the game under gles. It compiles and runs with regular opengl with all the new flags (including gles) activated.
I started working on the bump already you can follow my pull request here: https://github.com/gentoo/gentoo/pull/28 To sum it up, yes all the patches get dropped, the doc issue is fixed and gmp and libspatial are added as dependencies. I gonna look into luajit and gles.
(In reply to Martin-Kevin Neumann from comment #7) > > I gonna look into luajit and gles. I did my research now: opengles: As far as i can understand in this case, it seems that it only gets used by the android client of minetest and there is a upstream bug for missing transparency when using opengles2, so i am inclined to not activate opengles usage via useflags for now. luajit: The luajit useflag for minetest is masked by the gentoo devs, luajit itself is stable now and we got a stable version of minetest in the tree so i am unsure how relevant the comment in /usr/portage/profiles/base/package.use.mask is. Following the bugreport here: https://github.com/minetest/minetest/issues/2988, there is an unresolved memory issue that hits especially servers using luajit. Considering that i thinks it save to leave it masked for now.
Btw can the bug be assigned to me as proxy-maintainer?
I've been using luajit for the past couple months without issue. Nothing like the upstream out of memory issue you pointed out. I may go back to regular lua lua after seeing comments there about performance and maintenance issues. I just assumed luajit would be faster and haven't actually compared. Either way a masked use flag is no big deal. Regarding gles, I am hoping to put some time into running the game on Raspberry Pi at a playable framerate. _If_ I actually get around to it and get it working I'll bug you some more. Thanks for being on top of things!
Successfully bumped, for details look at: https://github.com/gentoo/gentoo/commit/3425759f0eea03b9034d2207a731ebed2548c573