BUG: pthread_mutex_lock on destroy

[pavlinux]

https://github.com/Feh/nocache/blob/2b6ea1f6b46dabd08db6c6b8be78874b90ecfd22/nocache.c#L207

May be unlock ? ))

[Feh]

No… you have to lock the mutexes so that you can guarantee no-one else is depending on them and the protected data being there, before freeing them…?

[pavlinux]

In general, the entire destroy() function is useless, except for two free()

[Feh]

destroy() and its attempt to clean up would be “useless” if the following conditions were true:

• Any FD that gets opened gets closed before program exit
• Any FD handling that the binary does go through libc syscall wrappers that nocache intercepts

For most even slightly more complicated programs, these conditions don’t generally hold.

You can see the reasoning in the original commit that introduced it: https://github.com/Feh/nocache/commit/d355dafdae2d382bf4af60252e53174718d73698

If you are worried about the constant overhead the destroy() function introduces, feel free to add some getenv() type call in there to optionally skip cleanup steps.