Files
Bubberstation/code/datums/elements/update_icon_blocker.dm
Jared-Fogle 45c14f6330 Adds SIGNAL_HANDLER and SIGNAL_HANDLER_DOES_SLEEP to prevent signal callbacks from blocking (#52761)
Adds SIGNAL_HANDLER, a macro that sets SHOULD_NOT_SLEEP(TRUE). This should ideally be required on all new signal callbacks.

Adds BLOCKING_SIGNAL_HANDLER, a macro that does nothing except symbolize "this is an older signal that didn't necessitate a code rewrite". It should not be allowed for new work.

This comes from discussion around #52735, which yields by calling input, and (though it sets the return type beforehand) will not properly return the flag to prevent attack from slapping.

To fix 60% of the yielding cases, WrapAdminProcCall no longer waits for another admin's proc call to finish. I'm not an admin, so I don't know how many behinds this has saved, but if this is problematic for admins I can just make it so that it lets you do it anyway. I'm not sure what the point of this babysitting was anyway.

Requested by @optimumtact.
Changelog

cl
admin: Calling a proc while another admin is calling one will no longer wait for the first to finish. You will simply just have to call it again.
/cl
2020-08-20 09:11:28 +12:00

13 lines
442 B
Plaintext

//Prevents calling anything in update_icon() like update_icon_state() or update_overlays()
/datum/element/update_icon_blocker/Attach(datum/target)
. = ..()
if(!istype(target, /atom))
return ELEMENT_INCOMPATIBLE
RegisterSignal(target, COMSIG_ATOM_UPDATE_ICON, .proc/block_update_icon)
/datum/element/update_icon_blocker/proc/block_update_icon()
SIGNAL_HANDLER
return COMSIG_ATOM_NO_UPDATE_ICON_STATE | COMSIG_ATOM_NO_UPDATE_OVERLAYS