(svn r3472) - [PBS] Remove from trunk. Anyone interested can still find it in branch/pbs. This reverts revisions r3158, r3140, r3075, r2977, r2674, r2625, r2621, r2529, r2528, r2525, r2524, r2519, r2517, r2516, r2507, r2499. (in conjunction with Tron)
- The only change is that the nsignalsw.grf file is kept and that existing nightlies with PBS signals get those signals converted to combo-signals.

(svn r3371) -Fix: don't allow road-vehicles to perform an out-of-bound operation while overtaking in a curve (tnx to guru3, Tron and peter1138)
-Codechange: while at that, removed a local variable (more pretty :))

(svn r3367) Unify the 4 distinct CMD_CHANGE_{AIRCRAFT,ROADVEH,SHIP,TRAIN}_SERVICE_INT commands into one CMD_CHANGE_SERVICE_INT command.
As side effect this is a
-Fix: The default AI tried to change the service intervals of vehicles via the CMD_CHANGE_TRAIN_SERVICE_INT command - regardless of the type of the vehicle - which of course failed for non-trains

(svn r3313) Remove GPMI related changes from trunk
Revisions in detail: 2542, 3226 (partial), 3229, 3231, 3232, 3238, 3242-3245, 3251, 3253, 3260, 3263, 3265, 3266, 3269, 3277, 3278, 3279, 3283 (partial), 3304, 3305, 3306

(svn r3232) -Add: implemented the event-system for AIs
-Add: added several hooks (event-callbacks) for road-related-stuff

(svn r3179) - RandomRange() and RandomTile() instead of home brewed versions
- CHANCE*() instead of mumbling strange numbers

(svn r3156) -Fix: removed some cases where autoreplace windows were redrawn when nothing was changed
due to the FOR_ALL_VEHICLES(), redrawing those windows are slow

(svn r3111) -Fix: [autoreplace] [ 1341783 ] Assertion failure in vehicle.c line 378
running MaybeReplaceVehicle() is now delayed until after the loop in CallVehicleTicks()
This avoids selling the vehicle the loop currently works with (and continues to work with afterwards)

(svn r3101) -Codechange: added _new_vehicle_id
this var works like _new_train_id and the rest of that kind of vars, except it is set each time a vehicle is build, nomatter what type
this is a nice tool to code vehicle independent code, which in turn can reduce code duplication
Right now it's used in ReplaceVehicle() and CmdCloneVehicle()

(svn r3081) -Codechange: [autoreplace] complete rewrite of autoreplace
this is a complete rewrite, that makes use of existing commands like build and sell
this means that multiheaded train engines are replaced correctly

(svn r3078) Some more stuff, which piled up:
- const, whitespace, indentation, bracing, GB/SB, pointless casts
- use the trinary operator where appropriate
- data types (uint[] -> AcceptedCargo, ...)
- if cascade -> switch
- if (ptr) -> if (ptr != NULL)
- DeMorgan's Law
- Fix some comments
- 0 -> '\0', change magic numbers to symbolic constants

(svn r2599) Fix: Road vehicle multistop handling used NPF even if NPF was off????
- Also simplified/optimized the code.
- Now it uses manhattan distance as an approximation instead of actual distance to find the distance. Much faster.

(svn r2516) - Feature: [pbs] Implement path-based-signalling. This allows multiple trains within the same signal block, provided their paths dont intersect. For this the block must have all exit and entry signals be pbs signals. Place these by ctrl-clicking 4 times on a normal signal.
- Feature: [pbs] Implement autoplacement of pbs blocks, when a block has an entry and an exit pbs signal, covert the entire block to pbs. Can be turned off in the patch settings.
- Feature: [pbs] Allow showing of reserved status by making the tracks darker, when the pbs debug level is at least 1.

(svn r2514) - Codechange: [NPF] Move the checking of railtype into a funciton IsCompatibleRail().
- Codechange: [NPF] Check the railtype along a route against the engine type instead of against the previouse tile. This clears the way for electriefied rails.
- Add: [NPF] [ 1209644 ] A penalty for crossings (peter1138)

(svn r2448) General cleanup of rail related code, more to follow.
* Add: rail.[ch] for rail-related enums and wrapper functions.
* Codechange: Removed dozens of magic numbers with below enums.
* Codechange: Rewrote CheckTrackCombination().
* Add: TILE_SIZE, TILE_PIXELS and TILE_HEIGHT constants.
* Add: enums RailTileType, RailTileSubtype, SignalType to mask against the map arrays.
* Add: enums Track, TrackBits, Trackdir, TrackdirBits for railway track data. (Note that the old RAIL_BIT constants are replaced by TRACK_BIT ones).
* Add: enums Direction and DiagDirection
* Codechange: Moved a bunch of track(dir) related lookup arrays from npf.[ch] to rail.[ch].
* Codechange: move RailType enum from tile.h to rail.h.
* Add: Wrapper functions for masking signal status in the map arrays: SignalAlongTrackdir, SignalAgainstTrackdir and SignalOnTrack.
* Add: Wrapper functions to access rail tiles, using above enums
* Add: Wrapper functions to modify tracks, trackdirs, directions, etc.
* Add: Wrapper functions for all lookup arrays in rail.[ch] (Arrays are still used in parts of the code)
* Codechange: Renamed some variables and arguments to better represent what they contain (railbit -> track, bits -> trackdirbits, etc.).
* Codechange: Don't use FindLandscapeHeight() in CmdRemoveSingleRail(), since it returns way too much info. Use GetTileSlope() instead.
* Codechange: [NPF] Removed some unused globals and code from npf.c.

(svn r2297) - CodeChange: server-check the next batch of commands.
- CodeChange: since only the server will be able to modify difficulty settings, leave the checking of correct values besides, and trust users will join legit servers.
- CodeChange: for renaming signs, only check if GetDParam(); eg _decode_parameters is empty ('\0') or not, instead of the extra check of players, etc. That basically does the same thing. Also dirty sign two times when renaming, once before, once after the action. Because if the name becomes shorter and you update only after, garbage remains on the screen.
- CodeChange: made GetMaskOfTownActions() available to the town-cmd to double-check if the action was available to the player. For this purpose the hardcoded _local_player has been removed from the function and is now passed as a parameter.

(svn r2262) - Fix: Assertion when vehicle in a depot wants to do pathfinding.
GetVehicleTrackdir now tries to get a valid trackdir as much as possibly, by assuming that a vehicle is facing outwards in a depot or road station, for example.
- Codechange: [Multistop] Multistop now also tries to find a slot for road vehicles that are in stations, since the pathfinder now properly handles that.

(svn r2261) - Fix: When crashed vehicles try to find a depot for servicing, openttd asserts.
Crashed vehicles shouldn't find depots anyway...

(svn r2256) - Fix: Trains cannot find a depot when they are in a tunnel. (glx)
- Add: GetVehicleTrackdir() helper function.
- Codechange: Moved SortStruct from vehicle_gui.h to ttd.h, so the dependency from vehicle.h on vehicle_gui.h could be removed.
- Codechange: Typedeffed the VehicleTypes struct so it can be used as the type for Vehicle.type instead of "byte".
- Codechange: Removed prototype for VehicleSorter(), which had no implementation anymore and was never called.

(svn r2024) -Fix: [autoreplace] reverted all changes involving v->set_for_replacement as they caused desyncs.

The bad sideeffect of this is that now no vehicle will automatically go to a depot anymore just
because it is set to be autoreplaced. We will have to find a better way to solve this problem.

Revisions reverted: 1640, 1707, 1709, 1710, 1712(but not the cheat prevention in this one)

(svn r1905) - Fix: [ 1118810 ] openttd: ship_cmd.c:642 ... Assertion failed. Mapwrap fixed in ship_cmd.c (was implicitely ok before biggermaps).
- CodeChange: rename all vehicle controllers to ...Controller for their similar behaviour

(svn r1863) Give the effect vehicle type enums more descriptive names and use the enum as parameter type for CreateEffectVehicle*()
-Fix: [1116619] Generate the correct smoke type for diesel trains

(svn r1834) - Fix: NPF does not check the owner of its target, busses try to enter other players' depots. TODO
- Add: asserts to find the v->u.rail.track == 0 problem.
- Add: IsValidDepot(), IsValidTown(), IsValidSign(), IsValidVehicle(), IsValidStation()
- Add: GetTileOwner(), IsTileOwner()
- Codechange: Replaced IsShipDepotTile(), IsTrainDepotTile(), IsRoadDepotTile() by IsTileDepotType().
- Codechange: typedeffed the MAP_OWNERS as Owner. Should be used as variable type.
- Codechange: Replaced a few uint by TileIndex.

(svn r1817) -Codechange: Moved depot-functions to depot.c
-Codechange: Added wrappers around depot-access (GetDepot no exists)
-Codechange: Made depot-functions a bit more logic (no longer
GetDepotByTile crashes your game when you request it on a non-depot tile)
-Add: made depots dynamic (yes, 64k depots are possible now)

(svn r1799) -Codechange: [ Multistop ] Added debug class "ms" and more debugging
output for multistops

(svn r1786) -Fix: unitnumber is increased to 16bit, so now you can have up to 5000
trains in one game (instead of the 240 which was the current value).
Default max allowed vehicles per type is changed:
Trains: 500 (old 80)
Road: 500 (old 80)
Ships: 200 (old 40)
Aicraft: 300 (old 50)
(Tnx to Celestar and Darkvater for checking the patch)

(svn r1770) -Fix: Hopefully last pieces of code that are containing a station-id
(and should be an uint16, not uint8)

(svn r1751) - Feature: New PathFinder (NPF).
- Supports trains, road vehicles and ships.
- Uses A* pathfinding (same codebase as the new ai).
- Currently unlimited search depth, so might perform badly on large maps/networks (especially ships).
- Will always find a route if there is one.
- Allows custom penalties for obstacles to be set in openttd.cfg (npf_ values).
- With NPF enabled, ships can have orders that are very far apart. Be careful, this will break (ships get lost) when the old pathfinder is used again.
- Feature: Disabling 90 degree turns for trains and ships.
- Requires NPF to be enabled.
- Ships and trains can no longer make weird 90 degree turns on tile borders.
- Codechange: Removed table/directions.h.
- table/directions.h contained ugly static tables but was included more than once. The tables, along with a few new ones are in npf.[ch] now. Better suggestions for a location?
- Fix: Binary heap in queue.c did not allocate enough space, resulting in a segfault.
- Codechange: Rewritten FindFirstBit2x64, added KillFirstBit2x64.
- Codechange: Introduced constant INVALID_TILE, to replace the usage of 0 as an invalid tile. Also replaces TILE_WRAPPED.
- Codechange: Moved TileAddWrap() to map.[ch]
- Add TileIndexDiffCByDir(), TileIndexDiffCByDir().
- Codechange: Moved IsTrainStationTile() to station.h
- Add: IsRoadStationTile() and GetRoadStationDir().

(svn r1741) - Fix: added IsVehicleIndex() so it's possible to protect GetVehicle() from reading an invalid vehicle index
- Fix: added check for v->type in some commands, which expects v to be a specific type

Checks like this is needed to protect network servers from people, who hack their clients to either cheat or crash the server

NOTE: if I made a mistake here it can make a function unreachable when it should be used. Here is one place to look if something weird happens

(svn r1735) -Fix: [ 1112285 ] Fixed an assertion error with stations where stops
were deleted, the assertions were just plain wrong at that place. Used a
more graceful handling. Todo: What happens if ALL stations in a schedule
have no suitable stop?

(svn r1721) -Feature: It is now possible to build multiple road stations (up to 8) on
a single station.
Thanks to: Truelight for the saveload code, Darkvater and Hackykid for
network testing and Tron for proof-reading 1500 lines of diff.

(svn r1652) Fix: train and road vehicle lists are now redrawn when vehicles enters/leaves a depot (Hackykid)
fix: the ship status bar was not updated properly after a breakdown if show_speed was off (Hackykid)

(svn r1629) added a counter to tell how many engines you have of each type to the autoreplace vehicle windows and made them show only the vehicles you actually have in the left list.
this also fixes some window updates issues when autoreplacing

(svn r1554) -Fix: [ 1103187 ] Order Check messages are now validated before
displayed, so that there are no stray error messages any more.
-Feature/Fix: Order Checking is only execute for ONE vehicle in an
order-share system

(svn r1525) -Codechange: rewrote the _order_array, now it can be made dynamic.
-Codechange: renamed all 'schedule' stuff to 'order(list)'
-Codechange: moved all order-stuff to order_cmd.c / order.h
-Codechange: vehicles that share orders are now linked to eachother
with next_shared/prev_shared in Vehicle

Developers: please use AssignOrder to assign data to an order. If not,
you _WILL_ make the save-routine to assert!

(svn r1407) -Codechange: changed a lot around _stations, _vehicles, _towns and _industries
(in prepare of dynamic arrays):
- DEREF_XXX is changed into GetXXX
- All direct call are directed via GetXXX
- struct Industry has now an index-field
- ENUM'd some stuff
- Replaced home built loops with FOR_ALL_XXX
- Added _stations_size, _vehicles_size, ... which gives the length of the
array (which will be dynamic in the near future)
- Changed lengtof(XXX) to _XXX_size (e.g. _stations_size)
- Removed all endof(XXX) (because mostly it was part of a FOR_ALL_XXX)
- Made the sort-functions of all 4 dynamic
- Made all 4 Initialize functions more of the same
- Some minor tab-fixing and stuff
(tnx to Tron for proof-reading my 100kb patch ;))

Note for all: please do NOT directly call _stations, _vehicles, _towns and
_industries, but use the right wrapper to access them. Thank you.
Ps: please also do not use 'v++', where v is of type Vehicle *.

(svn r1386) Move TileIndexDiff to map.h
Move _tileoffs_by_dir to map.[ch] and encapsulate it in TileOffsByDir()

(svn r1323) Adding autoreplace feature
This feature works much like autorenew, but it will get you a new engine type instead of a new one of the same type. Once
ordered, it will automatically replace the engines while they visits a depot. The GUI for setting this up have been added on the
vehicle overview windows
Note: autorenew is now autoreplace, but to the same engine type
Nice new features, that was added to make this possible
- windows can now have two independant vertical scrollbars
- CMD_SHOW_NO_ERROR have been added as a flag for DoCommandP. It will make it do the action instead of showing the red box with
estimated costs even if shift is pressed
- fixed problem where enginetypes where not initialized when loading a game. It's now done in InitializeGame()