Time  Nick      Message
07:15 nore      thoughts on #1119?
07:15 ShadowBot https://github.com/minetest/minetest/issues/1119 -- Fix rendering glitches when far from the center of the map by Novatux
12:20 sapier    did anyone ever mention android c library beeing a nightmare about temporarys and memory usage? if not now it's done
12:21 sfan5     one more reason to compile in a static libc
12:21 sapier    well that's exactly what I do
12:21 sfan5     which libc do you use?
12:21 sapier    still memory is messed up to an extent hard to believe it does even run
12:22 sapier    the one supplied with ndk
12:23 sapier    e.g. if I supply L"Some button text" to a button I see the text in gui, if I pass a pointer to allocated memory containing same text I don't see anything ... not even if I don't ever free that pointer
12:24 sapier    and to make it even worse this happens on real device only but not in emulator
12:38 sfan5     sapier: fix please, https://github.com/minetest/minetest/issues/1152
12:43 sapier    hmm that's strange sfan according to the testcode 2.0.2 is not denied
12:44 sapier    do you have multiple versions of luajit installed sfan5?
12:45 sapier    https://github.com/minetest/minetest/blob/master/src/CMakeLists.txt L231 I did only prevent some versions, unknown ones are allowed
12:45 sfan5     not that I know of
12:46 sfan5     I read the code
12:46 sapier    I don't see what could be wrong there, maybe you find the bug. You could try compiling that piece of code manually and see it's result
12:47 sfan5     $ dpkg -l | grep luajit | awk '{ print $2 }'
12:47 sfan5     libluajit-5.1-2:amd64
12:47 sfan5     libluajit-5.1-common
12:47 sfan5     libluajit-5.1-dev:amd64
12:47 Jordach   sapier, because the emulator uses the native C lib, instead of the Android one
12:48 sfan5     Jordach: how would ARM code use the native libc?
12:48 Jordach   sfan5, define emulator
12:48 sfan5     also it runs the whole linux kernel in qemu
12:50 sfan5     sapier: it works with gcc but not with clang
13:01 sapier    ok what does clang do?
13:04 sapier    well I expect android arm emulator to behave same as real device especially on STATIC linked lib
13:04 sapier    obviously I'm wrong about this
13:40 sfan5     sapier: seems like clang isn't the problem
13:41 sfan5     it fails with gcc randomly too
13:42 sapier    random errors?
13:44 sapier    I'd have expected code that simple to result in deterministic behaviour ;-)
15:47 kahrl     sapier: src/CMakeLists.txt:243, are you sure the sizeof does what you think it does?
15:47 kahrl     242*
16:40 sfan5     kahrl: correcting that does not help either, seems like cmake is broken
16:41 kahrl     weird
16:43 sfan5     my code: http://ix.io/aGt
16:43 sfan5     works with both clang + gcc when doing manually
16:44 sfan5     but it seems like cmake does not even get to run it
16:44 kahrl     sfan5: broken_luajit_versions is not an array
16:44 kahrl     it's a pointer
16:45 sfan5     it is used here too: https://github.com/minetest/minetest/blob/master/src/daynightratio.h#L45
16:46 kahrl     yes but "values" is an actual array so it works there
16:48 kahrl     ARRAYSIZE(broken_luajit_versions) is 1 on my gcc
16:48 kahrl     do this: static char* broken_luajit_versions[] = (char *[]) {
16:48 kahrl     then ARRAYSIZE(broken_luajit_versions) is 7
16:51 hmmmm     you don't need that cast at all btw
16:51 kahrl     right
16:51 hmmmm     why is it there in the first place
16:51 sfan5     still doesn't work
16:51 sfan5     hm
16:51 hmmmm     also how could sapier F something up so horribly
16:52 hmmmm     also in ARRAYSIZE you should have (a)[0], not a[0]
16:57 sfan5     I have http://ix.io/aGu now, how come it still says my lua is broken?
16:59 hmmmm     now, I don't know much about cmake, but from my understanding that takes the return code of that program and sets VALID_LUAJIT_VERSION to that
17:00 hmmmm     and you check if (VALID_LUAJIT_VERSION) right below, as in, if VALID_LUAJIT_VERSION is a non-zero value then do this
17:00 hmmmm     but a return code of 1 means it's an invalid version
17:00 hmmmm     I don't know if this is the problem or not but it's just something that looks fishy
17:02 kahrl     hmmmm: I checked /usr/share/cmake/Modules/CheckCSourceRuns.cmake, it compares the exit code to 0 and if so sets VAR to 1, and otherwise sets VAR to 0
17:02 hmmmm     oh.
17:02 hmmmm     hmmm.
17:02 kahrl     it's more complicated than that because of cross compiling so there might be a bug
17:02 kahrl     or something I overlooked
17:03 sfan5     nope
17:03 sfan5     switching the return codes around does not make it work
17:04 kahrl     what does if("${VAR}" MATCHES "^${VAR}$") do?
17:04 kahrl     it's a test in CHECK_C_SOURCE_RUNS that could prevent the program from being run
17:05 sfan5     it does not try to compile it in the first place when cross-compiling is used
17:13 kahrl     sfan5: shouldn't the elseif(CMAKE_CROSSCOMPILING) be else(CMAKE_CROSSCOMPILING)?
17:13 sfan5     yes it should
17:14 sfan5     ahhhhh
17:14 sfan5     works now
17:16 kahrl     it doesn't find my luajit.h for some reason
17:16 sfan5     mine is at /usr/include/luajit-2.0/luajit.h and it works just fine
17:17 kahrl     I have LUA_INCLUDE_DIR=/usr/include/luajit and luajit.h is in there
17:18 kahrl     oh wait, it's -2.0 as on your system
17:18 kahrl     but why doesn't it give an error earlier?
17:21 sfan5     https://github.com/minetest/minetest/pull/1156
17:22 kahrl     sfan5: looks good
17:25 kahrl     it seems my troubles with luajit.h were because of an outdated cmake cache, nevermind
17:28 sfan5     can I merge it?
17:30 kahrl     sfan5: yes
17:30 sfan5     done
19:09 nore      btw: did someone look at #1119?
19:09 ShadowBot https://github.com/minetest/minetest/issues/1119 -- Fix rendering glitches when far from the center of the map by Novatux