Game crashes on playing dynamic songs with only one section [linux, macos]

[samuelmarinsoto]

Describe the bug
Whenever the enemy brain is killed, the game will freeze for a few seconds and then crash.

To Reproduce
Steps to reproduce the behavior:

1. Start game
2. Go to scenarios
3. Start Bunker Breach scenario battle (happens on anything really)
4. Full gold, maximum AI intelligence for player and cpu
5. Start scenario
6. Kill enemy brain

Expected behavior
Win screen

Screenshots
If applicable, add screenshots to help explain your problem.

Additional context
here's the last error before crashing
/usr/include/c++/14/bits/stl_vector.h:1130: constexpr std::vector<_Tp, _Alloc>::reference std::vector<_Tp, _Alloc>::operator [with _Tp = unsigned int; _Alloc = std::allocator; reference = unsigned int&; size_type = long unsigned int]: Assertion '__n < this->size()' failed.
Aborted (core dumped)

I'm also running Nobara (Fedora based), compiled from latest develop branch (commit 34552e8)
github won't let me upload the coredump, so use this link: https://bin.dembased.xyz/corecortexcommand1000fa2728a4cbdc4f29bc28bd84d753688611174791732255150000000.zst
Ask me for it again if the link expires, or tell me how to upload it :)
Thanks!

[HeliumAnt]

Fixed in #162 ?
If you are using the flatpak release, it's already fixed! Otherwise if you're compiling yourself, apply this patch:
https://github.com/cortex-command-community/Cortex-Command-Community-Project/commit/1ef5eda87cfa714bbe13007c668298db25c52b7f.patch
Sorry missed that you're using current dev. Will need to check if that's still broken, but it does look identical to errors I was getting with error checking disabled

[michael-n0813]

Still having issues with game crashing when a brain is destroyed, with or without
luabind_args += ['-DLUABIND_NO_ERROR_CHECKING=1']

I am on Arch linux with AMD CPU and GPU.

[HeliumAnt]

Confirmed, game reliably crashes on loss or win, or if theres otherwise only a single song to select from.
@Causeless @pawnishoovy

[pawnishoovy]

yep, as said in discord i can't tell why this would happen. on Windows it appears to work fine. if we can't figure this out before release a dirty fix might be to insert a silent SoundContainer into single-container DynamicSongs so as to make them not single-container, which appears to be the problem if regular the battle music DynamicSong works fine.

something to test for mac/linux users:

does the single-container ambient music (appropriate activities call this when in initial edit mode i believe) also crash? does the first/some random SoundContainer from the battle music get eaten by any chance?

[michael-n0813]

I don't think this has anything to do with sound containers TBH.

• Ambient music work fine (single container)
• setting battle music DynamicSongSelection to a single sound container works fine
• adding more sound containers to victory/defeat still causes crashes
• adding a braincase in a custom scenario and starting scenario causes crash (unrelated to music)

Edit 1 actually I was wrong, got it to not crash modifing Music.ini
Edit 2 Hmmm need MORE testing

[michael-n0813]

OK so

• deleting everything from Music.ini causes game to crash on start
• deleting everything from Music.ini except for menu music and battle music works perfect
• adding more sound containers to single container components does not work

I think I have cracked it, having more than one dynamic song causes the game to crash.

[Causeless]

So my suspicion, given that it works on Windows but not Linux, is that we likely have undefined behaviour somewhere. I recall some iterations of the code around the dynamic music looks a bit wonky, I guess I'll need to look again.

Of particular note GCC has some interesting (stupid) ideas about certain cases of undefined behaviour, for example in functions without a base-case return it can literally just decide to literally omit the RET instruction.

[Causeless]

It's not undefined behaviour. It's literally just that libstdc++ is bounds-checking the vector whereas MSVC's std is not.
8993e9d probably get this to assert before crashing, but this is likely a data issue (and is likely resulting in some nasty ASAN issues even on Windows builds).

[michael-n0813]

So if I understand, it plays the DynamicSong (Battle Music for example) and upon calling a new Dynamic Song is removing (clearing) sound containers from the DynamicSong and then erroring out that the DynamicSong is empty?