polserver/docs/development-changes.txt

Development Changes - stuff that really doesn't belong in core-changes.txt
-------------------   (or, that does belong there, just not yet)

Format:
Date, Author, Files changed, Description of change
---
2013-09-02 Tomi
      Added Uobject->setmember<type>(member constant, value) and UObject->getmember<type>(member constant)
            These are for adding dynamic rw members stored in a map and using memory only when needed
	    These uses the constants from objmembers.h
	    Return value of a member defaults to 0 when it does not exist
            As an example UArmor .ar_mod is now changed to use this system

2010-07-09 Nando
      Fixed: strncpy(dst,src,n) does not always null-terminate the string dst. (in ssopt_parse_totalstat())

2009-08-18 MuadDib
    Changed: Core now disconnects all online clients to initiate the cleanup sequence after data is saved during
             shutdown. Should avoid any and all client/char/combat crashes during shutdown.

2009-08-03 MuadDib
    Note:    Remember, when working with adding new Message Handlers for incoming packets, if a packet has changed
             since the original (structure), there is a special way for the core to know this. You need to set up
             the new "struct" for it in the appropriate "pkt*.h" files, and for the MESSAGE_HANDLER line you need
             to use MESSAGE_HANDLER_V2() instead. This registeres the second structure type of the packet in the
             core's secondary Handler. See packet 0xB9 for one example and Drop Item packet for incoming.
             
2009-07-18 Turley
    Changed: Shadowrealms have now an ID saved inside a map<id,realm*> used for threadsafe identification of
             an deleted realm (decay)
    Note:    I think there is still a chance that decay thread can crash, maybe its better to only set a orphan flag
             on DeleteRealm and destroy the realm inside the decay thread

2009-06-26 Turley
    Fixed: Memoryleak inside datastore (no Deconstructors)
    Fixed: (no idea if really) linux crash if http thread couldnt start removed passert check

2009-05-26 Turley
    Fixed: (hopefully) crash on Shutdown inside schedule_attack() exit_signalled checks inside set_opponent()
           to skip the functions

2009-03-30 Nando
    Fixed:  Character::start_script() will now return false if the passert() would be triggered. No need to
            crash POL with a passert() if its recoverable. The passert() error was probably caused by the fact
            that spells are scheduled to happen always in 'now', which is useless and unnecessary. It only opens
            the possibility that the script will try to be run with another one already running... read to-do.txt
            for 099 to see what I mean.
            
2009-03-22 MuadDib
  Changed: void return_resources( u16 objtype, u16 amount ). Will now skip if exit_signalled is true. It was
           causing crashes during shutdown due to item cleanup, and it trying to produce more world resources
           for an item's cleanup after the ResourceDef was already emptied.
           
2009-03-05 Nando
    Added: Custom house now has the correct packet handlers.
    Added: Check on SendHousingTool() to check if the house is custom (must set with multi.setcustom(1))
     Note: The system is still very insecure, and should be made more customizable. Overall, what we had
           of code is working and bug free (so it seems). :)

2009-03-05 Nando
    Added: send_sysmessage_cl_affix(), say_above_cl_affix() and  private_say_above_cl_affix() to clfunc.cpp.
     Note: need to add the eScript functions in clemod.cpp  
             
2009-03-01 Nando
  Changed: Thread status report will now say how many child_threads and what's the size() of ThreadMap
           contents.

2009-02-25 MuadDib
  Changed: Rewrote transmit_to_inrange and transmit to others. They now use a bool of true/false
           to decide if this is going to a >= 6017 client.
  
2009-02-02 MuadDib
  * Added: Initial code for AOS Resistances. These are purely for storage, mods, and Statmsg
           handling right now. Still need to throw in checks in core cmbt and places that will
           need to know which to check for stuff like Statmsg does now.
           
2009-01-26 Nando
  * Added: Cleanup of Repsys Hooks
  * Changed: VersionClientStruct is now declared on the stack when needed, instead of the heap.
             (Like ConfigElem's) 
2009-01-26 Turley
  * Added: Cleanup of syshooks

2009-01-25 Turley
  * Added:
           (#define LEAKFINDER) - Added to xmain.cpp. Use this to detect memory leaks in
           the core. Remember, this must be uncommented, and compiled using Debug mode, not
           Release. :)
           
2009-01-25 MuadDib
  * Changed:
           is_banned_ips() no longer loads the bannedips.cfg file every client
           login attempt. Now it is loaded via new function in bannedips.cpp and
           a struct/vector holds the cfg info (created in bannedips.h that I added).
           It reloads it with reload configuration too now. Much cleaner.
           
2009-01-22 Turley
  * Fixed: 
           Cleanup of
           Spells::spellcircles & spells2
           UOSkills::uoskills
           PacketHookData::clean_packethooks() to call the deconstructor
           ItemDesc::objtype_byname & old_objtype_conversions
           Pkg::packages_byname
           Polwww::mime_types

2009-01-21 Turley
  * Fixed: Cleanup of 
           Attribute::attributes_byname & attributes
           BoatShapes::boatshapes
           CmdLevel::cmdlevels2
  * Added: added resourcedefs.clear() to Resource::clean_resources()
  * Changed: changed delete_all(vitals) to Vital::clean_vital() to also cleanup vitals_byname 
  * Note:  never try to delete a vector like *_byname which contains aliases results in an crash cause
           it tries to access bad pointer...

2009-01-20 Nando
	* Fixed: Leak in RegionGroup, was missing deleting all the Region's from regions_ vector. :)
	* Fixed: clean_resources() from resource.cpp was calling "delete[] (*itr).second", but
	         it should be "delete (*itr).second" as far as I know, as it's being allocated by
	         calling "new something" and not "new something[count]". :)
	         
2009-01-19 Nando
    * Fixed: unload_packages() wasn't being called in pol.cpp during cleanup.
    * Fixed: unload_aux_services() wasn't being called in pol.cpp during cleanup, nor was it prototyped
             there. One less leak. Thanks Turley for running the leak analyzer! ;)
             
2009-01-18 Nando
    * Added: clean_multidefs() will now run at the end of cleanup_vars().  
    * Fixed: 'vitals' wasn't being deleted. Added delete_all(vitals) to uvars.cpp.
    * Added: There was no ~Vital(). Added one to delete the exported functions.
    * Fixed: ~Realm() was crashing because of delete[] *zone instead of delete[] zone. :)
    * Fixed: the vector<Realm*>* Realms was leaking on shutdown. Added delete_all(*Realms) and delete Realms
             to the end of cleanup_vars().
    * Added: clean_multidefs() to multidef.cpp. Just need to find a good place to call it...
    * Note: anti-leak brigade began today.
    
2008-12-31 Nando
    * Fixed: Internal webserver's thread wasn't using pthread_exit / _endthreadex,
             and never called threadmap::Register() if WIN32 was defined. (Maybe 
             change www to use threadhelp::start_thread()?)
    * Note:  This fix will possibly give us information about zombie connections... *hopeful*
    * Added: Even though I think run_thread() with no args in threadhelp.cpp is never called, added
	         error printing when catching an exception.
	         
2008-12-16 MuadDib
    * Note: A Function that returns a BObject/BObjectImp that is not set to a variable
            or passed directly to a calling eScript causes a memory leak.
            aka, if not done like : return SetSize(newsize); // where SetSize returns BObjectImp
            instead done like : SetSize(newsize); // where SetSize returns BObjectImp
            
2007-03-08 Shinigami
    * Added pthread_exit and _endhreadex to handle thread closing as intended

2007-01-01 MuadDib
    * NOTE: In standingheight checks there is a nasty bug. Items NOT locked down
            (unmovable), are NOT being checked correctly to block movement.

2006-10-07 Shinigami
    * FreeBSD must use stdlib.h instead of malloc.h.
      Linux can use stdlib.h but malloc_usable_size is decleared in malloc.h only.
      Win can use stdlib.h but _HEAPINFO is decleared in malloc.h only.
    * FreeBSD needs Port libexecinfo to support backtrace functionality.
    * FreeBSD 6.1 doesn't know malloc_usable_size - Counter bytes_allocated disabled.
      Upcoming FreeBSD 7 will support it.
    * FreeBSD doesn't know libc-version.h - Version Check disabled.

2006-08-16 Austin
    * Improved error report message for armrzone.cfg on startup.
    * Moved the call to close the system tray to the top of the x_maininner() function in pol.cpp
      I called it with atexit() so any time POL does an assert, exit, etc. it will remove the icon, now.
      It still wont remove it if the user clicks the x button to close the console.. need to research
      windows events for that. 
      
2006-08-10 Austin
    * Startup data load changes:
      1. Fixed a crash that would occur when loading storage.txt if an item was omitted.
      2. If an item belongs in a container or on a character and neither are present and IgnoreLoadErrors
         is enabled, the item will be removed, rather than halting POL startup.
      3. If an item has an undefined objtype and IgnoreLoadErrors is enabled, POL will not halt on startup.
      
2006-06-23 MuadDib
    * Moved more of the tooltip code to a new tooltip.h/cpp file setup for better
      organization.
2006-06-17 Austin
    * BIG improvements to uoscrobj.cpp
      Will convert the string into a token equivalent then use that part of the code.
      This should cause a teensy speed-up as well as the elimination of a lot of redundant code.
      
2006-06-15 Austin
    * bool FindNpcTemplate( const char *template_name, ConfigFile& cf, ConfigElem& elem )
      This function is still used by stuff, so I moved it from textcmd.cpp to npctemplate.cpp
    * Improved code for GetConfigStringDictionary() to directly use String(string)
    * Removed core support for the 'showadv' privilege (not used anymore)
    
2006-06-10 MuadDib
    Removed Totally removed the need for useskill.h in core. It was used for depreciated
            stat gain functions.
            
2006-05-30 MuadDib
    Changed In spells.cpp, added support for the scroll objtypes of SE and ML spellbook
            types. However, in the "school" 3 entry, it is a copy of magery. This is just
            as a placeholder, because thus far no book can be found that uses the 301+
            spellid range on OSI.

2006-05-24 Shinigami
    Extended progverstr and buildtagstr upto 64 chars to hold Code names.

2005-12-05 MuadDib
    Changed How realms are handled with internal_MoveItem. Now you send the new realm
            to the function (updated Container move func also) instead of oldrealm. Oldrealm
            handling is done in the move function so as to properly replace the realm member
            if an item is not sucessfully moved.
    Changed How ~ItemGivenEvent treats the GivenTime Cprop in order for it to properly compare
            it to the given_time_ member.
    Changed Changed u8 initialization of client_lang member in the 0xBF packet handler to 
            char initialization for compatibility and ease of reading.

2005-11-26 Shinigami
    Changed "strcmp" into "stricmp" to suppress Script Errors.

2005-11-14 Austin
    Added internal_InBoxAreaChecks(x1, y1, z1, x2, y2, x2, realm); to uoemod.h and uoemod.cpp
          Function is responsible for adjusting coordinates to fit within the realm.

2005-10-17 Austin
    Added uoemod4.cpp
    Moved Move* uo.em functions to here.
    Moved internal_MoveItem() to uoemod4.cpp
    Added internal_MoveBoat, internal_MoveCharacter, internal_MoveContainer to uoemod4.cpp

2005-09-10 Austin
    Added to pkg.cpp GetPackageConfigPath(Package* pkg, string filename)
    Will use in the future to more easily check if a config should be read from
    pkgdir/config/config.cfg or /pkgdir/config.cfg. Too much code would need to be
    changed to make it work properly with cfgfile.em for 096 but for now it does help
    reduce some redundant code elsewhere. 

2005-08-26 Austin
    Added lastx and lasty setting at the bottom of char_select() in pol.cpp
    to fix a problem with realm switching after logging in.
    Renamed inform_logout to inform_leftarea for better reusablility.
    Renamed NpcPropagateLogOut to NpcPropagateLeftArea for better reusability.
    Added distance check to inform_leftarea.
    Added a NpcPropagateLeftArea line to uoemod2.cpp when a mobile changes realms.

2005-08-25 Austin
    Moved the pol_distance(x1, y1, x2, y2) function from npc.cpp to ufunc.cpp
    Moved its prototype from uoemod to ufunc.h
    Move was made so that it sits next to pol_distance(character, object) for easier
    function overloading.

2005-07-04 Shinigami
    Memory leak inside Script-Cache (scrstore::ScriptStorage)... (is not fixed, just detected)
    it's the same leak type than inside Read-/UnloadConfigFile() - but not critical for live shards

2005-07-01 Shinigami
    Memory leak inside Read-/UnloadConfigFile() is not really fixed!
    Core now never loads unmodified cfg files again into memory, and this will stop huge memory usage.
    Old, but still loaded cfg files are moved to pol/cfgrepos.cpp:oldcfgfiles.
    We have to think over... it's a problem with script-inside references...

2005-06-30 Shinigami
    Memory leak inside BLong found (most BLong's gets destroyed on Shutdown only).
    fixed_allocator is a little bit critical too, but the advantages (grouping smaller requests)
    are greater than the disadvantages (blocks don't get destroyed on shutdown).

2005-06-20 Shinigami
    A lot of Memory leak detection stuff added... just create a configuration using MEMORYLEAK
    as macro definition and recompile everything (Core compiled with these macro will show'em
    in start/pol.log/console too). To show usefull stuff, just call PolCore().internal(1)
    and take a look @ debug.log - just add this line to WorldSaver Cycle... it's very usefull
    to get memory statistics and was used to detect all three memory leaks above. 

2005-06-06 Shinigami
    linux compilation with -O3 (or -finline-functions (a part of -O3)) will destruct
    pathfinding code in some cases! server will crash... (tested with gcc 3.3.5)
    e.g.: FindPath(5393, 838, 45, 5394, 826, 60);

2003-07-27 Syzygy
    Changed boat code to use new map format

2003-06-18 Syzygy
    SaveWorldState( flags := 0 ), flags can be SAVE_FULL or SAVE_INCREMENTAL
    pol.cfg: ShutdownSaveType = full|incremental
    more AssertionFailureAction options: shutdown-save-full | shutdown-save-incremental

2003-06-08 Syzygy
    Offline PCs will once again be saved.

2003-05-19 Syzygy
    Added checkpoints to Character::die(), should help catch cases where a mob death doesn't completely kill the mob
       - the additional die() call should be removed, since the check is meant to avoid creating multiple corpses.
    Added npc_count to uobjcnt, and report in pol.cpp at exit

2003-05-18 Syzygy
    Converted a bunch of catch(...) to catch( exception& ).
    catching ... should be unnecessary now, since code no longer throws strings / char* etc.
    Changed modules RArray to use vector

2003-05-17 Syzygy
    Converted all the 'throw' statements that were throwing strings to instead throw runtime_error
       This will allow catch( std::exception& ) to catch everything thrown by the application

2003-05-13 Syzygy
    objecthash stores UObjectRef now, and there is a reaper thread that sweeps it occasionally.
2003-05-09 Syzygy
    Added clib/iohelp.cpp and clib/iohelp.h
    Started the objecthash storing UObjectRef instead of UObject*

2003-02-18 Racalac
    Added "localvarmembers" debugger command in poldbg.cpp
2003-02-14 Racalac
    Added "pc" debugger command in poldbg.cpp. not real useful to users until we release the debugger client

2003-01-12 Syzygy
    runecl: Added '-a' option to enable assertions (but Debug build only)
    runecl: defined INC_PASSERT=1 (debug build)

2003-01-07 Syzygy
	Set debug formats for all Release targets to PDB.  This doesn't seem to affect the size of the .exe image.
	This is in preparation for better crash dump analysis support

2003-01-01 Syzygy
	Changed all the .c files in clib/ to .cpp files
	New pol.cfg option: UsePolMapFile=[0,1]
		If true, uses files in realm/main/* for most map / statics / LOS info.
		If false, original UO data files are used.
		uoconvert.exe generates the POL data files:
			c:\pol> mkdir realm
			c:\pol> mkdir realm\main
			c:\pol> uoconvert.exe c:\uodata\