When done from plugins, it doesn't correctly initialize the vtable
because of some weird things MSVC does, so the only safe way is to
use df::allocate<df::foo>(). For consistency, it is also enforced
for code in the main library. It reveals the issue in the digging
invaders plugin, first found by warmist.
This change is linked to a modification in df-structures codegen.
The threshold is set at the level when they start to blink - normally
they would continue on with the job until they get a thirsty/hungry
thought, but immediately run off to eat if they lose the job (thus
refusing to load the engine after firing it). The code checks for
active sieges and whether there is a free replacement unit.
- fixed unregister to allow unregistering individual TICK events
- make registerTick return the resulting freq you have to put back in to the handler in order for it to be unregisterable
- when TICK events fire, they now automatically remove themselves from handlers[TICK], as they should
- make sure manageTickEvent fires once per tick
- redo tick event management to work with world->frame_counter
- the freq of a TICK EventHandler is now ignored
- unregistering for an individual TICK event is still broken
- you would have to know the modified freq to even begin removing it
- it is only removed from handleres[TICK], not from tickQueue
- redo unregisterAll to be more efficient
- use the map loaded/unloaded events instead of world loaded/unloaded for initializing/clearing event monitoring data
- get rid of silly unsigned variables
- handle loading and unloading of saves better
- made sure it never triggers bogus events on loading a save
- plugins should now handle loading themselves with onStateChange
- special case to make Buildings module keep track of all buildings correctly on load
- fixed syndrome detection
- was broken because of the switch from ticks-since-the-dawn-of-time to world->frame_counter
- added lastSyndromeTime global
- fixed possible off by one error when looking for new created items
- got rid of the unread lastTick global
- const'ed the function pointer array of eventManagers
Unlike changevein, which changes the material in an existing vein
object, with this feature you can set the material for individual
tiles, and the cache will write it out as vein objects.
The array inside the cache data is changed from layer material to
vein material, because layer material is easy to compute.
The mask argument of the Impl template is there because apparently
an inner template cannot be fully specialized, so there needs to be
some argument besides i.
- Linux makefile overrides and removes the optimization flags.
- Old linux gcc doesn't understand lambdas.
- MSVC doesn't like extern and explicit instantiation in same file.