When retro core uses environment_cb to get info about variable updates, it passes pointer to bool (1 byte) value. This pointer is then treated as pointer to gboolean (4 bytes) in retro-gtk/retro-environment.c and so its function get_variable_update overwrites not only expected, but also three following bytes of caller stack, causing everything to come down crashing.

Link for example:
This is caller - https://github.com/libretro/snes9x/blob/master/libretro/libretro.cpp#L951
and this get called - https://github.com/GNOME/retro-gtk/blob/f9f780192e5c5def12b31d782b9038e29aaa4dd4/retro-gtk/retro-environment.c#L212

Error & stack trace

*** stack smashing detected ***: <unknown> terminated
Process 11223 stopped
* thread #1, name = 'python2', stop reason = signal SIGABRT
    frame #0: 0x00007ffff76328a0 libc.so.6`__GI_raise + 272
libc.so.6`__GI_raise:
->  0x7ffff76328a0 <+272>: movq   0x108(%rsp), %rcx
    0x7ffff76328a8 <+280>: xorq   %fs:0x28, %rcx
    0x7ffff76328b1 <+289>: movl   %r8d, %eax
    0x7ffff76328b4 <+292>: jne    0x7ffff76328ce            ; <+318>
(lldb) bt
(lldb) bt
* thread #1, name = 'python2', stop reason = signal SIGABRT
  * frame #0: 0x00007ffff76328a0 libc.so.6`__GI_raise + 272
    frame #1: 0x00007ffff7633f09 libc.so.6`__GI_abort + 457
    frame #2: 0x00007ffff7675517 libc.so.6`__libc_message + 631
    frame #3: 0x00007ffff77056bf libc.so.6`__GI___fortify_fail_abort + 47
    frame #4: 0x00007ffff7705682 libc.so.6`__stack_chk_fail_local + 18
    frame #5: 0x00007fffdb5b9ad2 snes9x_libretro.so`::retro_run() at libretro.cpp:962
    frame #6: 0x00007fffe8227fae libretro-gtk.so.0`retro_core_run + 30


Changing argument type of get_variable_update to actual (bool*) or other 1 byte long value, such as (gchar*) fixes the problem.