1) Improved the road vehicle allocation (aka slotting) for multistop. Stops can now accept unlimited, er... 256, vehicles.
2) Removed the "wait for stop" feature, because it did not work in practise.
3) Slotting now ignores unreachable stations. Uses NPF at the moment because the old pathfinder cannot do it (yet)
4) Now matter how many vehicles approach a station, they will always be distributed evenly over existing stops.
5) Hopefully the last fundamental change to multistop
- CodeChange: Change the function GetCurrentLocale(). It returns the locale from some default environment-variables, plus a custom one defined as parameter. If all fail, it tries $LANG.
- Remove the temporary synchronisation in during the map-transfer as this is no longer needed
- The saved patches work just like the saved gameoptions. You have a _patches and a _patches_newgame struct. The _patches_newgame struct contains the values from the configuration file and thus the defaults for new games. When a new game is started or an older game is loaded, the default values are copied over to _patches to be used. When you load a game that has PATS saved, the default values are also loaded, but immediately overwritten by the values from the savegame. This ensures that player-based values are always taken from your personal preferences.
- The current implementation also changes the default values if you change player-based settings in the game. For example changing window_snap_radius in a certain game will also change it for all next OpenTTD sessions.
- The savegame version has been increased to 22.
- The last 6 orso patches close the following reports:
[ 1366446 ] different names for patches: all patch settings have the same name as in the configuration file and are reachable from the console.
[ 1288024 ] Strange string on OTTD initial screen: configuration (and this includes patches) inputs are validated and clamped to their minimum/maximum values.
[ 1423198 ] Make "Signals on Drive side" player, not server, based: this is only visual so current setting is to save it with the savegame but not synchronise in multiplayer.
[ 1208070 ] Patches and New GRF options saved: apart from newgrf this is done
- Code has been added to the saveload code to honour the SLF_SAVE_NO and SLF_NETWORK_NO flags. SLF_NETWORK_NO just reads in the the bytestream and then discards it because that setting is not synchronised. For this the function SlSkipBytes() has been reinstated
- SAVEGAME_VERSION has been changed from a constant ENUM to a constant integer. This was done for the configuration-code to be able to tell which version of a CONDVAR type to handle. As said before, because settings can be saved to the savegame, they will become conditional at some point. The configuration code always has to read the 'most recent' version.
- GameOptions are saved through the new structure. It is fully compatible with any old savegame...however it is better. Because of the move to this new format we can instruct the loader to skip certain variables. Autosave for example isn't synchronised anymore (in the network). The same goes for currency and kilometers :D. That is the only functionality change this patch is supposed to have if I have written it correctly.
- NOTE! Patches are still not saved so for Multiplayer to work network_client.c and network_server.c needed slight modifications.
- hide and validate variable-size calculation (SlCalcConvMemLen/SlCalcConvFileLen)
- some signed/unsigned fixes
- handle conditional variables the same way for global -and struct types, through checking version_from and version_to
- Change fixed array per player to a single pool. This avoids future problems
with vehicle numbers and decreases savegame size. Engine replacements from
previous savegames will be lost.
- Move engine replacement code from players.c to engine.c.
(thanks to blathijs for rewriting this)
- Fix: savegames only give back one message, show this by ignoring the first argument. Perhaps make the message more verbose in the future by adding STR_ equivalents next to the already existing debug messages.
This means that any user attempt to remove a rear engine will tell the user to move the front engine instead
This fixes the assert when moving multiheaded engines (introduced in r3144)
Note: to make old savegames use this feature, some engines might be turned around in order to link engines in pairs
-Codechange: train subtype is now a bitmask
This allows fast access to info like if it is a wagon or engine and if it is in front and so on
Note: savegame version bump
- Waypoints: Until now stat_id was saved but never assigned to. Instead we now save the GRFID/local index of the custom graphics so that GRF file changes can leave graphics intact.
-Trains will now remember the length of stations it visits and sell cars
when being autoreplaced if they became too long
-If it needs to remove cars, then it starts from the front and sells
all it can find until the train is short enough
-This only works for trains, that knows the station length of the route
so a full uninterrupted run is needed
-a train needs 1-2 runs to detect if the shortest station is expanded
-This feature can be turned on and off in the train replace window
and each company can have it's own setting
-NOTE: minor savegame version bump
When game is saved and you save again you get an error message (not when autosaving) and it's aborted. When a game is loaded in the meanwhile execution pauses until saving thread finishes.
-This means that one company can only have one setting for renew and replacing
more clients will not fight due to different settings anymore
-This is a needed step in the line to fix autoreplacing dualheaded locomotives
NOTE: savegame revision bump (peter1138 + me in coop)
- Do not dereference a local variable which no longer exists; this lead to random crashes when saving
- (Win32) Do not close a handle before it is used last
There are still many major problems (race conditions and resulting memory corruption/crashes) left
