- Split config cell_stack_limit into custom_cell_stack_limit (previous feature) and official_cell_stack_limit (see below)
- Expanded map_count_oncell by a flag parameter, currently only one flag is supported: only count standing units (needed for official cell stack feature)
- Added a new function map_closest_freecell that will return the closest free cell using the same order that official servers use
- Monsters will now actively search for a free cell when starting to walk randomly and when unlocking target
- When any unit finishes walking (regularly) and is not on a free cell, it will now actively search for a free cell
- Step actions will be delayed until a suitable cell was found, they will even be executed when the player walked slightly out of attack range
- Reverted all the icewall-related changes done in SVN r15777 and following as testing shows they aren't official and are actually pretty exploitable (bugreport:7412)
- Instead implemented the official icewall characteristic that monsters can only leave an icewall cell to the west or south, the changes include:
* The "sight" path check no longer checks for the current cell so standing on an icewall allows you to see/attack into any direction
* The path finding will still ignore the current cell as before but the walk routine will not allow to walk east or north while standing on an icewall cell
* This leads monsters in the situation where they go through an AI loop not allowing them to escape the icewall (if their target is north or east of them)
* Monster in this situation will use idle skills and if they get attacked will use their rudeattacked skills if available, similar to traps like Spiderweb
* Added a configuration icewall_walk_block that allows to configure how long a monster should go through the AI loop before the server allows it any movement, this "safety" system is official and seems to equal about 75 AI loops; if you want to disable the whole icewall system so that monsters don't get stuck in icewall at all, just set this to 0
* Here are videos from jRO showing how this system works: http://ragdo.blog56.fc2.com/blog-entry-763.html
- Implemented the official calculation for "direction"; now you will be considered horizontal/vertical/diagonally aligned with a target cell in the exact same way as on official servers, this is for example used to determine whether an icewall or a firewall should be horizontal, vertical or diagonal; the only thing that is still unofficial is the default direction (officially always "west"); effectively now there are more situations considered diagonal than before
- Fixed a compiler warning (converting double to float)
- Further cleanups on the idle skill use code for immobile monsters and monsters near a player but without a target (now skill using will always go via mob_unlocktarget)
* This also fixes that monsters switched to idle mode and start to use idle skills one second too late
- Added a new function unit_stop_stepaction which completely cleans up a planned action
- The new function will now be called when a new action request is received and when changing the map
- Removed setting of "chaserange" as part of the step action system as it's no longer used
- Changed check_distance functions to return "bool" rather than "int"
- Special thanks to Cydh for pointing out possible optimizations
- Removed redundant code
- Fixed two small bugs in the distance_client calculation
- Improved the "step action" system added in the last commit
* Added a "unit_step_timer" that is now used for all step actions; now even skills will always be executed at "full cell" (if you don't lag)
* The "step action" system will no longer mess with anything and rather just call the normal attack/skill functions that are also called by the client; this removes a lot of unneeded code and also works a lot better since the existing functions have been bug-tested the most; this also fixes a problem with the "auto-follow if you can't reach target" system
- Updated monster skill use behavior so monsters use skills the same way and at the same rhythm as official servers (bugreport:009326), the changes include:
* Unified the "monster can't move by default" and the "monster trapped" code as it really should behave exactly the same
* Fixed a bug that caused the "monster skill use" routine to be called 20 times more often than it should in some situations
* When a monster attacks you and you run away from it, the monster will now check if it has a ranged skill on "attack" state before switching to chase state
* Monsters will now always do a normal attack before using "attack" state skills
* Fixed a bug that caused monsters to switch to idle mode and never use their chase skills when they get hit continuously
* Changed default for skillrange_from_weapon from 30 (all but player) to 0 (none); monsters will now use all skills at the skill range and not at their own attack range, if they get tanked from farther away than skill range, they won't use these skills
- Updated ranges to work as on official servers (bugreport:009326), the changes include:
* Implemented new functions "check_distance_client" and "distance_client", that instead of the server-sided square range system use the circular one that the client uses; these functions should be used for units that send their commands via the client (players mostly)
* Applied the new distance algorithm to some player-specific range checks, players will now have a circular attack range that reflects their attack range on the client; this makes it impossible to hack the client for more diagonal range
* Removed the arbitrary +1 range bonus at range checks; as monsters now react fluently, they won't need it anymore, that means a monster with for example 4 attack range will now only attack targets within a 9x9 area around it, the moment you step out of this area, the attacks will stop (if the monster can move it will follow you); as for players, the extra cell attack range when on a linear line to the target now is integrated into the distance algorithm, that means a player with attack range 4 can attack from 5 cells away when directly on line with a monster, but no longer diagonally as range is circular for players now
* Implemented a new "step action" feature to reflect official chase range behavior; when an attack or skill is used inside the attack range, it will act exactly as now, the skill is used once the signal from the client was received; however, if an attack or skill is used outside the attack range, the player should move into the chase area which is 1 cell inside the attack range border; the client actually sends us where to use the attack / skill from, but previously we just dropped that information the moment an attack request came; now instead of stopping the player instantly on an attack request, the player will continue moving to the target cell and then automatically use the command received earlier (it will be remembered); this change was absolutely necessary as the client sends the attack request slightly before attack range is reached, execution on official servers only takes place on every full cell moved; the new system copies this behavior
- Implemented an improved hit-lock system (bugreport:007460), the changes include:
* MVPs are no longer immune to being stopped by a hit unless they used Endure
* When hitting any unit, it will stop for its "dMotion" interval, exactly at the end of "dMotion" it will continue walking (official behavior); this helps getting more "move" packets to be displayed
* The unit that was hit will be immune to being stopped for another "dMotion" interval, this allows to slow down any units by hitting them frequently, but makes it almost impossible to completely stop them forever (depends a lot of ASPD and dMotion value); this does not affect special hit-lock properties (some skills and events should set delay anyway)
* The unit that was hit will no longer be "pulled" to the next cell; this caused a lot of position lag, it should be much better now
- Fixed Sightrasher missing its variable cast time in Renewal
- This should solve most of the position lag problems, especially ones that caused monsters to suddenly appear next to you (happened especially often when using icewall)
- Added a new heap implementation to db.h (POP2, PUSH2, SIFTUP, SIFTDOWN, UPDATE) that simulates the heap that the client uses, there was no other way to reproduce 100% exact behavior
- I recommend using the old heap implementation for everything else
- Updated path.c to use the new heap macros and also fixed the order in which the different possible directions are pushed into heap
- Special thanks to rversteegen for helping me with various tests and the heap implementation
- Special thanks to ultramage for providing info that helped us narrowing down the possible variables
* Added and cleaned up comments.
* Added checks to prevent map server crashes.
* Cleaned up some reports from CPPCheck.
* Other various stylization cleanups.
(https://github.com/HerculesWS/Hercules/commit/78028c8)
* This fixes a problem where units sometimes walk diagonally through a wall server-sided; there should be less client desync now (but it might not be perfect yet)
* I kept the rAthena specific option path_blown_halt
* Special thanks to piotrhalaczkiewicz for the original implementation and aleos for helping me with the merge
* Fixed an operator check for WE_MALE and WE_FEMALE. Thanks Baalberith.
Optimizations
* Fixed a possible null pointer dereference in Spirit Ball check, Change Cart check, WE_BABY, MH_PYROCLASTIC, and MH_LIGHT_OF_REGENE.
* Fixed a reassigned value being reset before being checked.
* Fixed a homunculus expression always being false because of prior check.
* Fixed sprintf specifier output trying to display int value when return value is unsigned int.
* Fixed a redundant check for when monster's HP is less than 0 when HP is declared as an unsigned int.
* Fixed a redundant check for when uAccLen or uTokenLen are less then 0 when both are declared as unsigned int.
* Various other clean ups.
-Upd @mapinfo to count number of vendors in map
-Upd getrandgroupitem to check if valid itemit returned.
-Cleanup and KR style
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@17232 54d463be-8e91-2dee-dedb-b68131a5f0ec
-Change some remaining -1 for status ending timer.
-Move some hardcoded msg in msg_athena
-Small optimisation on battle_attr_fix
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@16835 54d463be-8e91-2dee-dedb-b68131a5f0ec
- It fixes issues caused by RAND_MAX being only 32k in Windows system (bugreport:1927, bugreport:86).
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@15483 54d463be-8e91-2dee-dedb-b68131a5f0ec
Also removed script object 'setcell', added script function 'setcell'.
- Now you can manipulate cell information without needing @loadnpc
- You can also manipulate the terrain ('gat') type itself, using the new cell_walkable, cell_shootable and cell_water constants
(currently the implementation uses bit flags too, so to get the type you want, you need to adjust the flags one by one)
- This breaks current scripts, so please adjust places that use setcell
(also be sure to _only_ use predefined constants, not direct numbers)
- Details can be found in the script reference.
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@12009 54d463be-8e91-2dee-dedb-b68131a5f0ec
- defined a data structure for map cells (replaces 3 various cell arrays)
- both terrain (gat) and dynamic (cell) information is now stored as C-style bit flags instead of #defines and bitmasks
- added map_gat2cell() and map_cell2gat() for terrain type conversions
- changing terrain information via 'setcell' is temporarily disabled
- mapserver startup now takes longer, as it needs to adapt mapcache data to internal representation, cell by cell (new mapcache format anyone?)
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@12003 54d463be-8e91-2dee-dedb-b68131a5f0ec
* Added map/path.h, moved path-related function headers to path.h.
* Removed the macroed _real() path functions.
* Modified some functions to use boolean return values instead of 1/0 or 0/-1.
* Modified path_search_long() to allow a NULL output pointer (in which case a temporary local buffer will be used instead).
* Removed an unused ->path_half member variable from struct walkpath_data.
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@11958 54d463be-8e91-2dee-dedb-b68131a5f0ec
* Added timestamps to the log of memory leaks.
* Moved definition of __func__ to cbasetypes.h.
* Configure script updated:
- added option to select the memory manager
- added option to enable MAPREGSQL
- added option to enable DEBUG
* common's Makefile deleting svnversion.h on 'clean' target.
(run ./configure again to update your Makefile)
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@11760 54d463be-8e91-2dee-dedb-b68131a5f0ec
- now works on occupied squares (previously it disappeared)
- now you can walk out of an icewall square (removed code that blocked pathfinding if the origin cell was of a nonwalkable type)
- added back the hack where mapcell changes are broadcast to whole map (prevents client from leaving the cells non-walkable if you warp away)
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@11354 54d463be-8e91-2dee-dedb-b68131a5f0ec
* Changed clif_sitting() to use 'bl' instead of 'sd' (for non-player objects)
* Removed way messed-up script function 'unitdeadsit'
* Tagged 'FIXME' lines written by myself
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@11040 54d463be-8e91-2dee-dedb-b68131a5f0ec
- added a missing \n to one login server warning
- silenced one "dc[4] might be uninitialized" warning
- fixed an apparently wrong abra_db.txt entry (venom splasher: 40000)
- removed .gat from the remaining npcs
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@10974 54d463be-8e91-2dee-dedb-b68131a5f0ec
- If jobchanging while disguised, the disguise will be removed, since it can't be kept and still calculate correctly the new job class to use.
- Fuuma shurikens are now unbreakable code-wise (like all maces, books, etc)
- Applied Meruru's patch to increase speed of easy path seeks.
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@9518 54d463be-8e91-2dee-dedb-b68131a5f0ec
- SL_STIN/SL_STUN have dex-reducable casting times now.
- Added a warning on item loading when an item is of equippable type but has no equippable position.
- Turning an alliance to an opposition won't work now during WoE
- Slaves can summon other mobs now. The only ones who can't do this are player summoned mobs.
- Merged the following functions for faster processing of items:
clif_storageitemlist + clif_storageequiplist = clif_storagelist
clif_guildstorageitemlist + clif_guildstorageequiplist = clif_guildstoragelist
clif_itemlist + clif_equiplist = clif_inventorylist
- Increased max knockback value support to 25.
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@7109 54d463be-8e91-2dee-dedb-b68131a5f0ec
- Added some missing SC_* entries on the initial listing (potion related and speed up ones)
- Fixed function declaration of map_getcellp()
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@6817 54d463be-8e91-2dee-dedb-b68131a5f0ec
- NPC break equipment stuff now has a 1.5% per skilllv success rate.
- pc_setpos won't random warp players if placed on top of Moonlight Petals
- Removed the check in clif.c for npc_shopid as the client never sends a packet when the trade is cancelled...
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@5384 54d463be-8e91-2dee-dedb-b68131a5f0ec
- Updated path.c with jA's implementation, which should make long path searching more efficient.
- Also added some typedefs from jA for the common structure types (PC/MOB/NPC, etc)
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@5366 54d463be-8e91-2dee-dedb-b68131a5f0ec
- Added flag 0x3000 to path_search to specify that path searching should go through stacked cells. Used for skill_wall_check calls.
- easy path_search will be successful in CELL_NOSTACK mode if the search fails only on the target cell (it is presumed the target cell contains the character one is trying to reach). The walk routines will take care of stopping the character from walking into another one when stacked.
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@5263 54d463be-8e91-2dee-dedb-b68131a5f0ec
IF YOU HAVE A WORKING AND TESTED BUGFIX PUT IT INTO STABLE AS WELL AS TRUNK. EVERYTHING ELSE
GOES INTO TRUNK AND WILL BE MERGED INTO STABLE BY VALARIS AND WIZPUTER. -- VALARIS
git-svn-id: https://svn.code.sf.net/p/rathena/svn/trunk@5094 54d463be-8e91-2dee-dedb-b68131a5f0ec
- Changed "read_map_from_bitmap" to "read_map_from_cache" in map_athena, "map_bitmap_path" to "map_cache_file"
- Fixed item effects not showing when only one was used
- Fixed a bug in Safety Wall
- Allow only either Storm Gust or Lord of Vermillion to cause damage if stacked together
- Added path_search_long, map_find_skill_unit_oncell
git-svn-id: https://svn.code.sf.net/p/rathena/svn/branches/stable@998 54d463be-8e91-2dee-dedb-b68131a5f0ec
