Doomsday Engine

2008-02-02

06:17 Revision 9b1535b6 (github): Reimplemented the sort algorithms in BSP_DetectOverlappingLines() and MPE_DetectDuplicateVertices() to use ptrs rather than indices. Plus a couple of other very minor tweaks.

danij

2008-02-01

09:16 Revision 8786f44f (github): Use a zblockset for allocating hedges when building the BSP. This has improved the speed of createInitialHEdges() by ~40%

danij

07:59 Revision c5ab4741 (github): Refactored the edgetip management in order to remove the need for a global index, which has now been removed.

Finished the editable map destructor. With luck, there are no remaining leaks. danij

07:24 Revision 31c57a15 (github): Moved hedge deallocation into the BSP destructor and other changes in order to factor out the need for a global hedge index, which has now been removed.

danij

06:45 Revision 92842116 (github): Replaced the algorithm for building the hedge sort buffer in buildSegsFromHEdges with one which traverses the BSP and builds the index as it goes. This will ultimately be a lot quicker as it means that we can use the spatial information to speed up the sort.

danij

06:20 Revision f8d8a0db (github): Use a local buffer when sorting hedges before hardening into the gamemap_t segs index. Once the hedge allocation is revised, this will allow for the current global index to be removed completely.

danij

01:58 Revision 28e6f009 (github): Revised how linedef flags are handled. Part 7 of 6 (oops, forgot about wolftc)

danij

01:57 Revision 5e8019ff (github): Revised how linedef flags are handled. Part 6 of 6

danij

01:56 Revision cb8fb380 (github): Revised how linedef flags are handled. Part 5 of 6

danij

01:55 Revision 35884fd6 (github): Revised how linedef flags are handled. Part 4 of 6

danij

01:54 Revision 63e77ffa (github): Revised how linedef flags are handled. Part 3 of 6

danij

01:53 Revision 80c3dae8 (github): Revised how linedef flags are handled. Part 2 of 6

danij

01:52 Revision b4e985b6 (github): Revised how linedef flags are handled. The game-specific linedef flags (read from an archived map) will be passed to the game via the map data custom property interface. The game can then interpret their meaning and use DMU to update the map accordingly.

Also, I have insulated the engine's linedef flags from those used by the engine by implementing a new set of flag def... danij

2008-01-31

20:10 Revision 1e9b10db (github): Use size_t for representing the length of HEdge lists.

Subsector creation moved out of edit_map.c and into the BSP code for now.
We no longer need global records of segs/su... danij

19:52 Revision 92781d4f (github): Added a DMU object selection modifier: DMU_SUBSECTOR_OF_SECTOR

danij

05:16 Revision 62a09944 (github): Plugged a memory leak in the destruction of the BSP structure in the node builder (all user data should be destroyed, not just that at leaves).

danij

04:41 Revision a8f16353 (github): Various enhancements to the BSP/nodebuilder:

* Extracted the data structure(s) used for the BSP tree and reimplemented using the new binary tree object. Note that... danij

04:29 Revision c5a1b12b (github): Implemented a fairly standard reusable binary tree object. Needs more robust error/usability checking and some unit tests (will add a later date). For now, take care not to do anything stupid!

danij

2008-01-28

03:10 Revision 2dbbd4d6 (github): Added execution timers to Sv_InitPools() and DAM_BuildBlockmap().

danij

03:08 Revision c2d85930 (github): Fixed: Attempt to free memory pointed to by the game-side things array when none have been allocated (wouldn't happen under normal use but still).

danij

03:01 Revision 1004b270 (github): Optimized the clockwise winding algorithm used in the creation of BSPs. By utilizing the newly created BSP structure itself for traversal it meant the global node and subsector arrays could be removed entirely.

Plus an additional optimization which uses a global hEdge sort buffer rather than allocate one every time it is neede... danij

2008-01-27

12:46 Revision a6cc5acd (github): Quick fix to prevent my bindings being nuked when running in -dedicated mode.

Strictly speaking there is no reason why bindings should be disabled when in -dedicated mode. Once we implement a con... danij

11:55 Revision b319938f (github): Fixed up our DirectInput keyboard use and the DIK_* to DDKEY_* translation. Plus other misc tweaks to the binding/input/event stuff.

danij

09:19 Revision 7cb4b6f0 (github): Moved the DirectInput keymap stuff out of dd_input.c and into win32/src/sys_input.c (disabled it for now also as it needs updating).

danij

08:19 Revision 863acdc9 (github): Pick number of visible whole lines + 1 when drawing the console (so that we draw the fractional part of the line that is not completely visible).

danij

07:48 Revision c45a6bff (github): Various fix ups to the window manager for all platforms.

danij

05:01 Revision b7b6a371 (github): Renamed global var trace > traceLOS to avoid symbol name conflict.

danij

04:41 Revision 967c6d82 (github): Added missing include for curses.h under *nix

danij

04:00 Revision e821dd4c (github): Management of console/terminal windows is now the responsibility of the window manager. Part 2/2

danij

03:58 Revision a3227910 (github): In order to reduce the likely-hood of -dedicated mode from routinely breaking; management of console/terminal windows is now the responsibility of the window manager and is interfaced through the same API.

Dedicated consoles/terminals now operate at keyevent_t level (rather than ddevent_t) and we now always allocate a ful... danij

2008-01-25

03:54 Revision 017b8a3d (github): Pass the cursor position to Sys_ConUpdateCmdLine instead of calling out to the console (cleaner).

Updated error messages in Sys_ConInit. danij

02:30 Revision e348087d (github): I have begun fixing up dedicated console mode under WIN32. It is once again functional but there is further work to be done.

danij

2008-01-24

13:17 Revision e24e6b15 (github): Ensure we have both key down and repeater bindings for the menu navigation keys.

danij

00:19 Revision decff35a (github): Added a _DEBUG check for attempting to write too large of a packet to a demo file.

danij

00:09 Revision b1a8ed31 (github): Use intptr_t for semaphore handles (as opposed to long) (64bit support).

danij

2008-01-23

22:57 Revision 89bca672 (github): These shouldn't be in svn.

danij

22:40 Revision f6b8af01 (github): Use time_t instead of long when representing system date/time in finddata_t (more portable).

danij

22:30 Revision 89bf7fd9 (github): Revert to rev 5298 (prior to the importation of the unit testing and float comparison stuff from deng-ng).

danij