structs inside their array, and possibly invalidating pointers higher up.
Meaning that any function called within an wndproc could cause unknown/invalid pointers
once control was returned to this function. Solved by the introduction of an extra
abstraction layer, an array of z-window positions that is only concerned with the
pointers.
WWT_IMGBTN must contain an image for drawing. Renamed WWT_PANEL_2 to WWT_IMGBTN_2
because that is what it is. Added WWT_PUSHBTN that is either just a pushable button,
or a textbutton, which text's drawn dynamically independent of widget.
Moved the actual modification of railtype to WE_INVALIDATE_DATA in the train depot handler
-Codechange: added SetWindowDirty() to WE_INVALIDATE_DATA as it made no sense to update the list without making the window dirty
Since they were freed with the rest of the array, it only meant that we wasted a few bytes (max 16) while the window were open and we didn't leak memory
-Fix: when retiring an engine design, invalidate the build windows and invalidate the build window data
-Fix: mark build windows dirty when engine reliability changes
Example: make a train transport iron ore from A to B, then it visits a depot and refits to steel
It then transport steel back to A or near A if there is a factory and then it visits another depot to refit to iron ore again
This is controlled in the orders. If a goto depot order is lightlighted, then "Unload" changes to "Refit"
Control click "Refit" removes the refit part of the order (as the tooltip says)
The player will still pay the normal refit costs
Known issues:
If a vehicle is not in a depot, then the refit window will fail to tell refitted cargo capacity
Refit costs in the refit window can sometimes print 0 when it should not because the refit calculation is unaware that the vehicle will be refitted in between
Warning: autoreplace got a protection against replacing something so you get a new cargo type, but it can fail here. In the iron ore/steel example, it can see that
the vehicle carries iron ore and the new one can be refitted to iron ore, then it will replace. It will not check to see that it's valid for steel as well.
This is something to look into in the future
It made no sense to maintain 8 nearly identically arrays when a single one can do the job
Also made the two buttons always use half of the bottom width each, even when resizing
This change is intended to make it easier to make depot behaviour consistent
and faster to code when adding more features in the future
The user interface should hopefully not be affected by this
This will ensure that you can always get the same list when checking for vehicles in a depot (no need to duplicate code for each place, that needs such a list)
Since the vehicles are only looped once for each redraw, drawing speed is around twice as fast (measured to be 114%-121% faster depending on the number of vehicles in the game)
This turned out to be due to OFB_HALT_IN_DEPOT and OFB_SERVICE_IF_NEEDED using the same bit
It appears that it doesn't matter for the code, so I adapted the string selection code to handle this
DestinationID being a union of these types is just hassle without benefit and cannot be handled correctly everywhere because of local lack of information
-Codechange: unified the code for mass goto depot to avoid duplicated code
-Fix: Vehicles already on the way to depots will not be cancelled by mass goto depot (made it really hard to send all vehicles at once)
