Re: [AD] SF.net SVN: alleg:[11915] allegro/branches/4.9/src

[Milan Mimica <milan.mimica@xxxxxxxxxx>]

Peter Wang wrote:
> Yes, I see.  Not all of the examples crash, for some reason.

The only one not crashing are ex_vsync and ex_draw because they call that 
function conditionally.

> This wouldn't solve the problem for new threads though.  What do you
> think about implicitly calling al_reset_new_display_options() in
> al_set_new_display_option(), if it wasn't explicitly called by the user?

This didn't work because some functions like al_get_num_display_modes() were 
calling _al_get_new_display_settings()... But I'm not sure this is necessary any 
more with the new API that was added.

In gerenal, If we do require users to init TLS in each thread they are going to 
use allegro in, then we should require them to explicitly call 
al_reset_new_display_options().

Probably a much better solution is to make ALLEGRO_EXTRA_DISPLAY_SETTINGS in TLS 
static. This would solve the leak too.


--
Milan Mimica
http://sparklet.sf.net