Activity - Doomsday Engine - Tracker

Activity

From 2008-03-09 to 2008-03-22

2008-03-22

22:05 Revision b0f13796 (github): Fixed z offset bug with mobj-linked light halos.: danij
22:04 Revision 0912f31a (github): Fixed clipping issues with decor light halos.: danij
18:46 Revision e194a9d9 (github): Implemented a series of mechanisms to handle the transfer of game-specific map object data types and their values into the games. This code is early and there is a lot of room for optimization but svn is once again playable using jDoom (yay!). The rest of the supported games will be updated shortly.: danij

2008-03-21

15:51 Revision b885f805 (github): Separated out the model skin management from the model code itself in preparation for wrapping with material_t.: The engine-internal data pointers in the shared state_t have been removed (they are supposed to have no meaning outsi... danij
00:18 Revision 3910e684 (github): Use the material_t interface for deleting and precaching of sprite textures.: danij

2008-03-20

23:57 Revision 735c4d50 (github): Wrap references to sprite textures (non-translated, not psprites) with the new material_t abstract interface.: Fixed a performance bottle neck in R_AllocRendPoly() (there remains more work to be done. I have to wonder whether in... danij
17:47 Revision 72cdf8c0 (github): Introduced new type; typedef int lumpnum_t - To be used at all times when representing indices to lumps (and lump infos).: Further cleanup, got rid of some garbage in dd_main.c danij
15:11 Revision 58b9075e (github): Re-jigged the DENG_LIBRARY_DIR stuff under Windows a bit. Now accepts both relative (to the location of Doomsday.exe) and absolute paths.: danij
14:07 Revision 24b42cca (github): Determine global paths (userdir, basedir etc) as early as possible during startup, so that under Windows we can provide LoadLibrary() with absolute paths. Also added DENG_LIBRARY_DIR which if defined at compile time; changes the location of the library dir under Windows relative to the base directory (Added for dev purposes only. It should NOT be defined in release builds (a compile time warning will be generated if so)).: danij
14:01 Revision 6ca8083e (github): Don't include the WolfTC project when building by default.: danij
01:28 Revision 894bf243 (github): Revert the trunk to rev #5514. What I'm intending to do is too intrusive for trunk so instead, I'll use branch/danij-experimental.: danij

2008-03-19

02:58 Revision 3181f10f (github): Revert changes from rev #5519. Evidently, SDL is not willing/able to cooperate with my plan of doing user input in a plugin. Instead, I'm going to do something radical; a plugin based arch in parallel with a "SDL mode" which removes all the abstractions and reverts to handling input in the engine and only ever one window.: danij
00:53 Revision ef866e91 (github): Split off our interface to SDL input out to disdlinput.dll/so/lib and added a di* lib loader under *nix.: danij

2008-03-18

23:54 Revision f06bff6d (github): Don't link Doomsday against dinput8.lib under Windows, as it is no longer dependent.: danij
23:40 Revision f1356c6d (github): Began splitting up the user input system. Part 3 of 3: danij
23:39 Revision 2d502d54 (github): Began splitting up the user input system. Part 2 of 3: danij
23:36 Revision 8e52b6ee (github): Began splitting up the user input system. OS and/or api specific input device interfacing code is being moved out of the engine an into "Input Driver" plugins. Internally, Doomsday has only one, platform independent representation for each type of supported input device. It is the job of the Input Driver plugins(s) to provide input event notification in a standardized format which the engine uses to update it's internal virtual devices.: Under Windows, we now have diDInput8 which is a Doomsday Input Driver that interfaces with DirectInput8. Next up, wil... danij
15:26 Revision b482fd85 (github): Ensure our window class' menu name is initialized to NULL, make sure we specify the window type we want when calling Sys_CreateWindow().: danij
15:24 Revision be269f44 (github): Fixed mismatch between the declaration and implementation of Sys_CreateWindow().: danij
01:05 Revision 5a7551fe (github): Fixed a bunch of compilation issues.: danij

2008-03-17

07:00 Revision fdf54bf3 (github): Removed old source file references from codeblocks/doomsday.cbp: danij
06:46 Revision fc9f8af4 (github): Added inter project dependencies to the codeblocks doomsday.workspace to ensure correct build order.: danij
06:38 Revision 8a57d300 (github): Added a virtual build target "All" for all projects. This build target will automatically build all child targets for each project (for example, if the active project is jDoom; Build "All" will build both Debug Win32 and Release Win32 targets).: danij
06:08 Revision 1bb5ec92 (github): Fixed relative paths in the codeblocks project files. Changed doomsday.lib include path to include the output target folder for each target build (e.g. search Debug Win32 when linking a Debug Win32 build of jDoom).: I can now build and run Doomsday using the Visual C++2005/2008 compiler from within codeblocks and use EDITANDCONTINU... danij
04:52 Revision 4b1fc175 (github): Ignore *.layout: danij
04:49 Revision 345dea49 (github): Evidently the *.layout files are not part of the project configuration.: danij
03:31 Revision 71288473 (github): Initial commit of a Code::Blocks workspace and projects for Doomsday and all its plugins.: danij

2008-03-16

21:30 Revision d0921ec5 (github): Revised the interface to GL_PrepareMaterial() now takes a material_t ptr rather than a type and index. Also, register DDTEX textures as materials so they can be referenced using the same interface.: danij
21:24 Revision 58e02668 (github): Reworked things so that we no longer have to call R_UpdateSurface() after any changes.: danij
21:16 Revision 4c18f787 (github): Added new public functions, removed old ones not currently in use.: danij
21:15 Revision ed48b000 (github): Revised DMU constant names.: danij
21:14 Revision 47149dab (github): Added new source files to doomsday_cl.rsp: danij
21:11 Revision bea0962f (github): Added new source files to the vs vcproj and sln. Also added the missing lib libcmt to the includes for a release build.: danij
21:04 Revision 12a2c6a5 (github): Worked on DMU:: * Split up DMU in to multiple source files. The public interface and backend stuff is still in p_dmu.c. The set/get f... danij
20:50 Revision efdcf7d6 (github): Updated DMU constant names.: danij
06:17 Revision 8afc3c3c (github): Implemented the remainder of the P_FindSector* utilities.: danij
05:16 Revision 0e55f728 (github): Fixed a bug that somehow crept into findBuildNeighbor().: danij
05:11 Revision d0258069 (github): Reimplemented XS_TravBuildStairs(). Rather than combining two quite different traversals into one routine they have instead been split into separate subroutines. Each subroutine was then rewritten to make use of P_Iteratep().: danij
03:04 Revision 597680f9 (github): Got rid of the global XG sector builderdata flags array. The new scheme isn't as optimal (requiring an iteration of the world sectors to zero the flags before a stair build begins, rather than a simple memset) but imo, it really doesn't matter.: danij
02:32 Revision 95285d22 (github): Got rid of XS_AdjoiningPlanes() completely. Instead, use the P_Iteratep() based search utility functions in p_mapspec.c: danij
01:48 Revision 83cc976a (github): Replaced the algorithm used in XS_GetPlane() to determine an extremal texture height of all the wall surfaces in a given sector. The new algorithm makes use of P_Iteratep() and unlike the old algorithm; it avoids the need for a static working buffer (which it never checked the bounds of, meaning that if a sector was made up of more than 128 linedefs and each had a different texture; it would result in writing beyond the end of the working buffer).: danij
01:42 Revision c8ba5209 (github): Polyobjs are no longer managed in the same way. Determining of their mirror will need updating (commented out for now).: danij

2008-03-15

23:29 Revision 034bec19 (github): Replaced recursiveSound() with a version that makes use of P_Iteratep. Also, moved it to p_mapspec.c as it is common between all supported games.: danij
22:31 Revision 770dc4e3 (github): Replaced EV_DoDonut with a version that makes use of P_Iteratep.: danij
11:02 Revision e5dcdcb3 (github): Don't forget to actually return the height of the next highest floor in P_FindSectorSurroundingNextHighestFloor - (crikey, what a mouthful...): danij
10:58 Revision f5ae6c94 (github): Updated more game-side algorithms to make use of P_Iteratep.: danij
09:44 Revision 687d4569 (github): Rewrote most of the game-side algorithms that require iteration of a sector's linedefs to take advantage of P_Iteratep.: danij
06:24 Revision 9daa1364 (github): Take advantage of DMU's new P_Iteratep() when drawing the automap. This will greatly reduce the amount of time DMU has to spend determining types, checking for dummies etc, etc. The net result being that the drawing of the automap causes less slowdown.: danij
04:28 Revision f7600a87 (github): Fixed a bug hat crept in in the previous commit.: danij
04:14 Revision 7ddc9c2b (github): Several improvements to DMU:: * Got rid of the unnecessary complication of runtime aliases for DMU_ property constants in the engine and replaced t... danij

2008-03-14

03:25 Revision ef493b3f (github): Simplified the use of R_GetMaterialColor(): danij
03:05 Revision 19c61e60 (github): Added more error checking to the MPE interface.: danij
02:41 Revision 4de4512f (github): Implemented MPE_VertexCreatev so that many vertexes can be created in one go, rather than having to call MPE_VertexCreate repeatedly. Made us of this in dpwadmaploader to speed up transfer of map vertex data to the engine.: danij
01:56 Revision 716f245e (github): Always clear both the color and depth buffers on the first frame after map load.: danij
01:06 Revision 571e4695 (github): Replaced triangulateSubSector() Due to the BSP leaf generation algorithm ensuring that a subsector's segs will always be in clockwise order it meant that there was no need to sort the segs in triangulateSubSector() and as a result, a nice little speed boost in map load times.: danij

2008-03-13

23:42 Revision 442c1e0a (github): Removed the automatic conversion of large lumobjs to bias sources. It wasn't working particularly well and will be replaced.: Fixed a bug in prepareSubsectorForBias() which failed to initialize the seg illum array correctly. danij
05:58 Revision 12c10dce (github): Variable name conventions, cleanup.: danij
05:16 Revision b483cb93 (github): Cleanup.: danij
05:13 Revision 522c0b2f (github): Fixed a couple of comments.: danij
05:11 Revision 5678d042 (github): Continued work towards the $nplanes goal. Begun removing the distinction between floor/ceiling planes and generalizing so that we use the plane normal instead.: danij

2008-03-11

04:46 Revision c4792b9e (github): Moved the utility routine isNonGlowingSurface to r_world.c Plus renamed it as R_IsNonGlowingPlane due to it being for planes and not generalized for any surface.: danij
04:37 Revision a66685fc (github): Clean up.: danij
04:35 Revision 8d8a8ca7 (github): Fixed a bug in the R_Find*LineNeighbor lineowner traversal routines that would result in an incorrect angle delta being returned when traversal is done anticlockwise.: danij
04:30 Revision 85753e0e (github): Fixed a bug in the lineowner ring, relative angle calculation. The last line in the ring ended up with an incorrect angle.: danij
00:33 Revision d50d1dc1 (github): Fixed a bunch more compiler warnings and one error reported by gcc.: danij

2008-03-10

23:27 Revision 9d200d21 (github): Fixed a bunch of warnings produced by gcc in the portable SDL window manager. Note, this code remains untested.: danij
23:06 Revision 3111adcb (github): Fixed some compile "errors" reported by gcc.: danij
08:25 Revision 5073d5a6 (github): Protected a bunch of calls to M_Free in sys_filein.c: danij
07:06 Revision 65764140 (github): Fixed a nasty bug in B_DestroyAllClasses() which could allow anyone to easily crash their system via the console.: danij
06:12 Revision 3940854e (github): Completely revised the implementation of fakeradio shadow edges. The old scheme did not take advantage of the spacial information provided by the vertex lineowner rings particularly well.: Todo: No method for resolving overlaps has been implemented as yet. Only the vertex inner offset is actually used. Ne... danij

2008-03-09

11:56 Revision 4825c4f8 (github): Optimized Rend_RadioSubsectorEdges() a bit.: danij
10:26 Revision 7801fda4 (github): Moving the fakeradio shadow poly edge offsets into lineowner_t provided the possibility to get rid of shadowpoly_t completely as the other properties it held were actually inferred by the structure(s) it linked to. Also, revised R_ResolveOverlaps to get rid of the requirement of an array to sort within (now uses a linked list).: danij
06:04 Revision 0d7a2515 (github): Fixed: Fairly frequently, the vertex coordinates generated for fakeradio edge shadowpolys would result in "folds" at the inner offset. This was due to us storing two separate offsets for each side of the edge and due to numerical inaccuracies; these would sometimes end up not being equal. This has been fixed by using only one set of offsets for each pair of edges and storing them in a the vertex lineowner ring for easy access.: Todo: This has the shortcoming that all edge pairs need to have a set of shadow offsets allocated for them as opposed... danij
05:15 Revision aeaf9650 (github): Revised the shadowpoly_t structure in preparation for splitting the data off into the vertex lineowner_t structure.: danij
05:12 Revision 819d4659 (github): Added the option for specifying a different public name for an internal data structure in makedmt.py: danij
05:01 Revision 55f919c0 (github): Determine the inner and outer coordinates for fakeradio shadow edge polys before going into radioAddShadowEdge(): danij

Also available in: Atom