• Ivan Trombley (8/8) Dec 27 2017 I believe that I should call SDL_Quit before my program
• Mike Parker (14/22) Dec 28 2017 No. GC cleanup happens after static destructors are run.

Ivan Trombley <itrombley dot-borg.org> writes:

I believe that I should call SDL_Quit before my program 
terminates. However, some objects may be holding on to SDL 
resources and freeing these resources (particularly truetype 
resources) causes crashiness if SDL_Quit is called before. Is 
there a way to have SDL_Quit (and TTF_Quit) called after the GC 
has completed shutting down on program exit?

If not then it could make things very complex and messy to keep 
track of these resources.

Mike Parker <aldacron gmail.com> writes:

On Thursday, 28 December 2017 at 06:34:00 UTC, Ivan Trombley 
wrote:
 I believe that I should call SDL_Quit before my program 
 terminates. However, some objects may be holding on to SDL 
 resources and freeing these resources (particularly truetype 
 resources) causes crashiness if SDL_Quit is called before. Is 
 there a way to have SDL_Quit (and TTF_Quit) called after the GC 
 has completed shutting down on program exit?

No. GC cleanup happens after static destructors are run.

 If not then it could make things very complex and messy to keep 
 track of these resources.

You really, really, really shouldn't be depending on class 
destructors to free resources like that. There is no guarantee a 
class destructor will ever be called and they are 
non-deterministic, so you can never ensure they will be called in 
a non-problematic sequence.

There are a number of alternatives here. The one I recommend is 
to use structs to wrap your SDL resources. It's what I settled on 
long ago and works well for me. Yes, you will have to take a 
couple of extra steps to cover all your bases if you don't want 
to manually release things, but it's possible to keep it simple 
and avoid messiness.