Time  Nick        Message
00:33 sofar       I'll take a look at the PR
00:43 erle        sofar at which one? the one from fleckenstein or from luatic?
09:51 rubenwardy  yay
09:51 rubenwardy  I can confirm that Google Play is giving nice tracebacks now
09:52 rubenwardy  first crash is a main menu issue - "no future without main menu"
09:52 rubenwardy  we don't have access to debug.txt so there's no way to know the Lua error
10:32 celeron55   wait where does that error text come from
10:33 MTDiscord   <paradust> oh, I remember encountering a main menu problem. Forgot to look into it. But if you create a world without specifying which game, a big Lua error popped up
10:33 celeron55   why can't the lua error end up in the same place
10:34 rubenwardy  that the error message comes from reading the source code
10:34 rubenwardy  it calls `abort();` so Google Play just has a SIGABT with a line
10:35 rubenwardy  (GUIEngine::GUIEngine()) (SourceCode: /home/ruben/dev/minetest_android_release/android/native/jni/../../../src/gui/guiEngine.cpp:202)
10:35 rubenwardy  specifically
10:43 celeron55   would be nice to be able to include an error message somehow
11:38 erle        celeron55, i do not know too much about C++, but can you just catch SIGABRT?
11:38 erle        void die(int s) { std::cout << "[last lines of debug.txt here]\n"; exit(s); }
11:38 erle        int main () { signal(SIGABRT, die); while(1) {}; return 0; }
11:38 erle        i tried this just now, but maybe it does not work on android for some reason?
11:40 erle        ok i checked, this also works if the abort() call comes from inside the house
11:44 sfan5       ...catch the abort to print an useful error message that we could just have printed directly?
11:45 sfan5       the issue is that google play crash reporting only has a backtrace but no way to include an error message or log lines (afaik)
11:46 erle        the thing is, i bet abort() is used more often than once. also i do not know if you need special handling for android error messages.
11:50 sfan5       the backtrace can already tell us where abort() was called
11:56 celeron55   well, can we catch abort and then somehow cause a backtrace with more information? 8)
11:57 celeron55   make function for each alphabet that can be called using an array of letters and which will recursively create a stack with the given letters and then actually abort
11:58 celeron55   if it works, it's not stupid
11:59 MTDiscord   <savilli> make sure it won't be inlined
12:01 MTDiscord   <savilli> but src/gui/guiEngine.cpp:202 doesn't have a error message, what do you want to print here?
12:26 celeron55   i guess it gets kind of complicated as the actual problem has probably occurred some time earlier
12:29 MTDiscord   <savilli> you can try to print the whole debug.txt that way, but it looks quite ambitious
12:38 celeron55   if we absolutely need a crash report sending function and a server for receiving them, it's not impossible to do it, but for more than a decade we've managed without
12:47 rubenwardy  "Managed without" depends on your tolerance for bugs
12:47 MTDiscord   <savilli> well, for more than a decade minetest has a poor android support
12:47 rubenwardy  yeah
12:48 rubenwardy  better support for error reporting on Android would be nice as a start, wouldn't be against extending it to other platforms but Android+GooglePlay is the ideal situation for it
12:49 rubenwardy  as 1) it's currently obscure to debug on android (you can't just ask users to grab a debug.txt)
12:51 rubenwardy  2) was going to be privacy, but that's not actually as a huge difference as you'd still want a method to opt out
12:53 MTDiscord   <savilli> why can't you ask users to grab their debug.txt?
12:53 MTDiscord   <savilli> I feel that automatic crash reporting should be out-in, reports can contain a lot of private data
12:54 rubenwardy  it's complicated
12:55 rubenwardy  currently, you often need to install a new file browser to access it
12:55 rubenwardy  we could add an option to share the file with a share intent, but that would only work if the app isn't crashing
12:55 rubenwardy  actually, I'll do that now
12:56 rubenwardy  I do opt-out automatic crash reporting in my apps, I'm very careful to never log any private data. But it's easier when you do this from the start and don't have modding support
12:57 celeron55   i guess there could be an android launcher that handles any crashes
12:58 rubenwardy  the Java is effectively already an Android launcher
12:58 rubenwardy  you could add some UI there to handle it crashing
12:59 celeron55   displaying the last lines from debug.txt in a copyable element would make sense
13:25 paradust    A pattern I've seen (and like) is having a BaseException class that stores a full stack trace at the time it is created (using the backtrace API), and stores it along with the message
13:25 paradust    And a top-level handler that catches all exceptions and logs them regardless of the logging levels
13:29 rubenwardy  #12370
13:29 ShadowBot   https://github.com/minetest/minetest/issues/12370 -- Add support for sharing debug.txt on Android by rubenwardy
13:33 MTDiscord   <MisterE> That is... very useful. You could eg text it to yourself.
17:45 rubenwardy  another android crash https://gist.github.com/rubenwardy/8d3383974edc959f21fa14e31706fc36
17:46 rubenwardy  https://github.com/minetest/minetest/blob/master/src/debug.cpp#L74
17:47 rubenwardy  https://github.com/minetest/minetest/blob/stable-5/src/script/cpp_api/s_async.cpp#L209
17:47 rubenwardy  Seems to be related, another mainmenu issue
17:48 MTDiscord   <savilli> another lua error without context ?
17:48 rubenwardy  yeah, it's the same phone model
17:48 MTDiscord   <savilli> how much RAM does it have?
17:48 rubenwardy  Redmi 9T
17:48 rubenwardy  3.8GB
17:48 MTDiscord   <savilli> should be enough
17:49 MTDiscord   <savilli> maybe they're trying to run a huge modpack
17:49 MTDiscord   <savilli> I wonder how OOM would look for android
17:50 rubenwardy  it's in the main menu
17:50 rubenwardy  both errors are in the main menu
17:51 MTDiscord   <savilli> ah, indeed
17:54 MTDiscord   <savilli> it can be literally anything
17:54 MTDiscord   <savilli> buggy CPU, corrupted flash storage, bit flip during the installation
17:55 rubenwardy  I suspect it can't read the mainmenu builtin files
17:55 rubenwardy  either a permissions issue or something wrong with the installation process
17:55 rubenwardy  perhaps they're low on storage, and it didn't install correctly
17:55 rubenwardy  I don't think we handle the low storage case at all
17:56 MTDiscord   <savilli> does it unpack itself during the installation or smth?
17:56 rubenwardy  yeah
17:58 MTDiscord   <savilli> https://github.com/minetest/minetest/blob/master/android/app/src/main/java/net/minetest/minetest/UnzipService.java there?
17:59 rubenwardy  yeah
18:00 MTDiscord   <ROllerozxa> the fact minetest android even needs to extract game data like that into the user's data directory is kinda dirty. like, it probably should be like a RUN_IN_PLACE=0 build where builtin, fonts, minetest_game and other stuff like that is kept inside the app's assets folder instead of spilling over
18:01 MTDiscord   <savilli> I guess you can't save your world to the app's assets folder
18:01 rubenwardy  the share folder can be separate to user data folder
18:01 rubenwardy  which is the case when MT is installed to /usr
18:01 rubenwardy  share is /usr/share/minetest (builtin, games/minetest_game)   and user data is ~/.minetest
18:02 rubenwardy  but yeah, ideally it wouldn't have both a .zip in the app and an extracted version on disk
18:04 MTDiscord   <ROllerozxa> yeah so stuff like worlds and user-installed games and mods would still be in the data folder
18:05 MTDiscord   <savilli> looks like it's handled https://github.com/minetest/minetest/blob/51cfb57b4da8e71e39380bf5bf3e19414c779e12/android/app/src/main/java/net/minetest/minetest/MainActivity.java#L72
18:18 MTDiscord   <savilli> oh lel, I've found the problem
18:18 MTDiscord   <savilli> it won't start the unpack next time on this condition https://github.com/minetest/minetest/blob/6901c5fae54eafb05494823b60d4e26c14b342f1/android/app/src/main/java/net/minetest/minetest/Utils.java#L31
18:19 MTDiscord   <savilli> ofc even if the directories exist, it doesn't guarantee that all the files were unpacked
18:22 rubenwardy  it's based on both version code and that function
18:22 rubenwardy  https://github.com/minetest/minetest/blob/master/android/app/src/main/java/net/minetest/minetest/MainActivity.java#L155
18:22 rubenwardy  version code is updated with any release
18:22 definitelya Could a core-dev open an issue on GitHub for this, please? (Wuzzy asked me to propose this there)   https://forum.minetest.net/viewtopic.php?f=5&p=410567#p410567
18:23 MTDiscord   <savilli> what's sharedPreferences? maybe they uninstalled it and installed again?
18:25 rubenwardy  shared preferences is an Android API a bit like get_mod_storage
18:25 rubenwardy  if they reinstalled then shared preferences will be reset
18:29 MTDiscord   <savilli> no idea then
18:33 MTDiscord   <savilli> is external storage smth like sd card?
18:41 rubenwardy  No
18:41 rubenwardy  It used to be, but it's just another storage location internal to the device
18:41 rubenwardy  Welcome to the Android sdk
18:41 rubenwardy  Full of random legacy stuff like that
18:58 sfan5       ROllerozxa app assets are not a folder or we would have surely done that
18:59 sfan5       they stay inside the apk on disk and you can read them using some java apis, no use for Minetest
19:08 MTDiscord   <ROllerozxa> so I assume the android asset reader inside of irrlicht doesn't work then?
19:10 sfan5       it probably does, but how many places have you seen inside Minetest that access the file system through irrlicht?
19:10 rubenwardy  yeah, we don't want to use more Irrlicht stuff
19:10 rubenwardy  ideally, the Android app wrapper would just provide a File URI to the C++ to access as a normal filesystem
19:11 rubenwardy  this probably could just work already, but it's still a .zip file
19:13 MTDiscord   <ROllerozxa> oh my bad, I thought minetest made use of irrlicht's file IO code
19:14 MTDiscord   <ROllerozxa> ...wait, minetest doesn't use any of irrlicht's file IO code?
19:16 rubenwardy  there's probably some exceptions
19:16 MTDiscord   <ROllerozxa> yeah only thing it uses is for extracting zip files
20:35 sfan5       merging #12343, #12329, #12325, #12321, #12368 in 10m
20:36 ShadowBot   https://github.com/minetest/minetest/issues/12343 -- Docs: clarify spawn_by by Zughy
20:36 ShadowBot   https://github.com/minetest/minetest/issues/12329 -- Fix no_texture.png being shown (instead of unknown_node.png) for unknown nodes w/ content ID < 125 by Wuzzy2
20:36 ShadowBot   https://github.com/minetest/minetest/issues/12325 -- Add missing concurrency protection in logger by paradust7
20:36 ShadowBot   https://github.com/minetest/minetest/issues/12321 -- unordered_map instead of map for MapSectors by RichardTry
20:36 ShadowBot   https://github.com/minetest/minetest/issues/12368 -- [no squash] Fix hash implementation for SerializedBlockCache by sfan5
21:09 MTDiscord   <x2048> Pushing #12357 in 5m
21:09 ShadowBot   https://github.com/minetest/minetest/issues/12357 -- Quantize light frustum calculations by x2048