For SPEEDYNESS
It probably works. I mean it's hard to sit there for 8 hours and test it. But it seems to work okay. At worst if it doesn't you'll have weather that doesn't make sense. ;v
* Added "Display Initialize() Log" admin debug command so you can see it mid-round.
* Ported the core of the overlays management subsystem from /tg
- Added SSoverlays subsystem for compiling overlay lists and applying them to atoms in a controlled anti-lag subsystem.
- Added vars and procs to atom which should eventually replace all direct interaction with BYOND's /atom/overlays var outside the subsystem.
- Added OVERLAY_QUEUED flag to var/atom/flags bitfield.
- Added small framework for subsystem performance tracking. So far used only by SSoverlays
- Added admin debug command "Display overlay Log" to see performance stats mid-round.
* Fix runtime on universal pipe adaptor update_icons
* Workaround for appearance_bro not initialized
Unfortuantely BYOND's initialization order is strange, and the appearance_bro var is only half initialized when map starts to load, causing errors. We temporarily fix by moving it to be a global-scoped global.
* Convert fire alarms to use add_overlay() A good first test.
* Convert turfs to use add_overlays(), eliminating the turf_overlay_holder!
- Converted as much as I could find about turf overlays to use add_overlay().
- This should be enough to stop BYOND from crashing, so we can eliminate the turf_overlay_holder hack.
- This also lets us remove the anti-corruption hacks from walls and open space.
- ZAS gas overlays can use priority overlays, so this also fixes the gas-goes-away-when-crowbarring-plating issue.
- Stuff like that
* Convert turf overlay interactions to use add_overlay.
Note: This is a plain and simple conversion of existing code to use SSoverlays. However I look at the line changed, and note that that line likely never fully worked as intended, as it has no way of re-applying itself.
I would make it use a priority overlay, but there is no code present for *removing* said overlay from neighbors when it is no longer required. That code should be implemented by original author.
* Move the global spawning_turfs lists from master_controller into SSxenoarch. Rename all references.
* Move the call to SetupXenoarch from master_controller.setup() to SSxenoarch.Initialize() Put SSxenoarch init order near the end to match current behavior.
* Replaces the shuttle_controller and shuttle process with the shuttles subsystem. Instead of docking ports being initialized by the game ticker, its part of the StonedMC Master init order.
* The main advantage of this is control over the initialization order, as well as letting Master be aware of CPU we're using up with shuttle processing.
* By being part of the Master init order, we reduce the uncertainty about "are objects initialized yet?" which is nice, since shuttle docks break if machines aren't finished initializing!
* Moves proc/initialize() from being on /atom/movable, /are and /turf/simulated to being on /atom - Now turfs can initialize too
* Added the SSatoms subsystem which controls initialization of atoms at roundstart and during normal conditions.
* Disabled the old auto_init = 0 behavior, ALL atoms should get initialized() called on them now.
* Refactored the way initialize() is called during /New() to utilize SSatoms instead of SScreation
* Removed SScreation, as it was only a stop-gap until SSatoms could be ported.
* Updated the maploader to inform SSatoms when it is loading maps instead of SScreation.
* Updated the template map loader to use SSatoms to perform initTemplateBounds
* Renamed 'initialized' var in seed_storage to deconflict.
* Removed usage of auto_init = 0, replaced with a no-op initialize() proc for atoms that don't need initialization.
* Creates the SSair subsystem which replaces the Setup, Start(), and Tick() procs of air_master.
* It may be best to have SSair completely replace air_master, but for now we are having them work together. It does completely replace the old processScheduer air ticker however.
* Remove the obsolete Setup and Tick procs.
* Adjust admin and debug verbs that dealt with some ZAS internals to work with the new implementation.
* This is PHASE 1 of a multi-phase conversion. In this first phase we implement the subsystem, but leave it processing the existing global list variables. In the next phase we will switch to use datum variables in the subsystem. The main reason for splitting into two phases is ease of code review; change the meaningful code without the hundreds of machines -> SSmachines.machinery substitutions.
* We did declare macros for adding/removing things to the processing lists, and convert everywhere to use the macros.
* Added var/is_processing to /datum to keep track of whether an instance is already in a processing list (prevents it being in the list twice!) and also debugging, making sure its not in two lists etc.
* NOTE: The global machines list is **no longer sorted** for performance reasons. As far as I know, the only module that actually ever cared was cameras. Our camera system already handles its own sorting in the cameranets anyway, so it should no longer be needed.
New lighting system is needed to handle updating literally 14k+ tiles whenever the sun moves. The straight port seems to have had no adverse effects from what I could see. Don't know if it's more performant but it doesn't seem to be less.
I didn't expect the latter would be needed for the former.
