AddressSanitzer et al. exit leaves terminal echo disabled.

[AliceLR]

When xmp is terminated by AddressSanitizer, MemorySanitizer, etc. encountering an error and exiting, terminal echo is still turned off until manually fixed with stty echo, tput reset, or similar. I don't know what the correct solution is for this, but I tried adding an atexit hook to call reset_tty and it didn't help. This has been mildly annoying when diagnosing issues like libxmp/libxmp#319 with input files that cause a lot of ASan/MSan exits.

[debrouxl]

Based on what some other projects do, I'd suggest:

• either adding a special fuzzing-oriented mode to the existing xmp, enabled by a #define: reduced initialization (e.g. hard-coded null device), reduced / eliminated terminal setup, etc.
• or probably better - slightly faster execution, no xmp source code clutter - making a second fuzzing-oriented libxmp front-end: a standalone, file-oriented (xmp_load_module_from_file), possibly slightly less bare-bones variant of the libfuzzer-based front-end I posted at OSS-Fuzz and/or libFuzzer integration. libxmp#302 - maybe basic module information could still be displayed upon successfully loading the module.

[sezero]

Anyone has a patch for this?

[AliceLR]

Not yet. I think the first one wouldn't be all that much clutter to xmp, and ideally it'd just suppress the player display portion of the terminal output (which is presumably the reason echo/etc is changed). The driver portion could just be handled with the driver flag, if needed (WAV output or playback might be desirable in some cases). It'd also be good to have a way to target any arbitrary one of the four xmp_load_module* functions.

edit: -q suppresses the player portion but it also suppresses the other output (which might not matter much if libxmp is built with -DDEBUG).

[sezero]

Maybe use a santizer callbak, e.g. __sanitizer_set_death_callback ??
https://stackoverflow.com/questions/47983039/clang-sanitizer-callback-on-error

[sezero]

It'd be nice if we can resolve this for 4.3.1

[AliceLR]

I've also found similar issues happen that require a tput reset, e.g. if you Ctrl+C mpv while a video is playing the cursor remains hidden, so I'm not sure how unusual this actually is.