Rip out the pcx loading code and use LoadPCX instead.
pcx.h:
remove the data field from pcx_t: it was messing with gcc's idea of the
structure size. also, update LoadPCX's prototype
pcx.c:
change LoadPCX to return tex_t (new type in texture.h) and take a new
param: convert. LoadPCX will now wither load raw 8 bit or convert to
rgba on loading depending on convert. Also, make LoadPCX WORK and use
Hunk_TempAlloc to store the resulting texture.
texture.h:
define tex_t. defines an in-memory texture (either 8 or 32 bit, depending
on the presense of a palette). Data comes immediately after the sturcture.
eg tex_t *tex = Hunk_TempAlloc (sizeof (tex_t) + count)
mismatch. First, the md4 checksum is stored in the mesh, and the md4 checksum
of the mesh and the model checksum are also stored. If, on loading, either
checksum fails, the model is re-meshed.
include/mdfour.h:
Add define for MDFOUR_DIGEST_BYTES
source/.gdbinit
setrom _windowed_mouse 0 so mods don't make debugging difficult
source/model_alias.c:
pass original model data and size to GL_MakeAliasModelDisplayLists
source/sw_model_alias.c:
GL_MakeAliasModelDisplayLists take (but ignore) new params
source/gl_mesh.c:
do md4 checksums on the model and mesh
qfgl_ext.h:
Add protection against multiple includes of extensions from different files.
Add defs for GL_EXT_paletted_texture.
Change the #ifndef conditional for shared palettes to
GL_EXT_shared_texture_palette.
addition, removed support (for now) for SGIS and EXT multitexture. It
allowed me to enhance support for ARB multitexture, which is better anyway
(especially for Radeon, but that won't get special support yet until we
need a third texture pass).
Updated the Unix 8-bit GL code to be more like the Windows version, which
may work better. It still defaults off, because many drivers have bugs in
their handling of the shared texture palette extension.
strings (skybox and *progs). skybox is easy to take care of (aditional parse
functions), but I haven't yet decided what to do with *progs. Also, progs code
needs a bit of a cleanup, but that will come later.
progs engine no longer refers to any global vars. This alows for multiple progs
being loaded (NOTE: NOT multiple mods, but (eg) for various protocol
extensions) and, once the last couple of issues are taken care of, CSQC.
Added editloc to change the description of the closest location
Added delloc to remove closest location
Added dumploc to place locs in memory into a .loc file (required for
editloc and delloc)
Added zdumploc to place locs in memory into a .loc.gz file (requred for
markloc, editloc and delloc)
Added loc.gz saving support to markloc, editloc and delloc (via
zdumploc)
Altered locs_load to use _Com_FOpenFile instead of Com_FOpenFile
Fixed potentual bug in locs_markloc
Fixed bug in Team_ParseSay that cause wierd behaviour if $ or % was last
charactor in line.
patch created from a development tree via diff -ur ../newtree ./
newtree and development tree were up-to-date at time of creating the
patch
Please let me know what you change so I may learn from it
Chris Ison (WildCode)
another alpha blend, it's set up exactly as another factor in determining
the final blend. This should help speed up framerates a little.
New cvars:
cl_cshift_bonus: Set to 0 to disable bonus flashes.
cl_cshift_contents: Set to 0 to disable content blends.
cl_cshift_damage: Set to 0 to disable damage blends.
cl_cshift_powerup: Set to 0 to disable powerup (quad, pent, etc) blends.
All of these cvars work for both GL and software, and all of them check
the value of the "cshifts" serverinfo. cshifts is a bit field, with the
following bits defined:
bonus: 1
contents: 2
damage: 4
powerup: 8
An admin can force any of these to be respected by choosing the numbers of
the cshifts to be enforced, and adding them up.
examples:
serverinfo cshifts 15
turns them all on.
serverinfo cshifts 10
turns on powerup and contents shifts.
also, the gl_cshiftpercent Cvar no longer does anything. I'll remove it
entirely soon.
rest: Myriad stuff.
r_netgraph now recognizes and abides by the cl_hudswap Cvar.
show_fps and show_time are on the same line now, and stay out of each
others' way.
New Draw_ function: Draw_TextBox(). It's like M_DrawTextBox(), except it
doesn't screw around with the screen coordinates.
rest: Implement model animation interpolation for alias models. Trying to
implement this has been an annoyance for some time, but IT WORKS NOW. set
gl_lerp_anims to zero to disable.
new function WriteTGAfile. Also, put in the tga header structure.
NOTE: this is currently broken for non gcc ocmilers. you will get a
#error message.
include/quakefs.h source/quakefs.c:
new function COM_WriteBuffers. Like COM_WriteFile, but scatter/gather
style.
source/gl_screen.c source/screen.c:
misc cleanups. differences are getting fewer. should be mostly
mergable soon (minus truly sw/gl specific code).
interface. To detect an extension, call QFGL_ExtensionPresent(). To get
the address of an extension function, call QFGL_ExtensionAddress(). This
function uses the preferred method for finding a function, based on
whatever OS and windowing system you use.
(for sv_minqfversion) are no longer compared as floats, so a server admin
can require client versions with a lot more granularity. (e.g. 0.2.99beta5
is greater than 0.2.99beta4, but lower than 0.3.0)
And of course the very slight rework of the input system, at least
in_x11 gets shifts right for different keymaps.
Oh, by the way, ALL other input targets are broken in a very 'will not
compile' way.
cl_main.c: Clean up version string a little
cmd.c: Add setrom parsing to first init phase, clean up includes a little
cvar.c: setrom command now uses set's semantics. Cvar_Set, Cvar_Set_f,
Cvar_Setrom_f now do a Con_DPrintf to report attempts to set CVAR_ROM
cvars.
sv_main.c: Clean up version string a little
vid.c: -winsize, -width, and -height no longer use Cvar_SetROM. The code
looks a little uglier, but it's more correct.
