Optimizations are left to the reader :P. I would prefer to look for ways of
giving the compilers hints about loops to unroll instead of having to read
the resulting mess of manual unrolling, though.
git-svn-id: https://svn.eduke32.com/eduke32@4728 1a8010ca-5511-0410-912e-c29ae57300e0
On Linux, it started manifesting itself since r4719 as crashes whose
backtraces pointed to malloc() and were not helpful. Valgrind finally
uncovered this.
git-svn-id: https://svn.eduke32.com/eduke32@4724 1a8010ca-5511-0410-912e-c29ae57300e0
The winding of a loop -- with clockdir() -- is determined by examining the
two line segments spanned between the points following a leftmost point of
the loop. If the loop contains a leftmost point that belongs to the "right"
side (as can happen with sliding door constructions), there's a chance that
an outer loop is misclassified.
git-svn-id: https://svn.eduke32.com/eduke32@4580 1a8010ca-5511-0410-912e-c29ae57300e0
And use these in jmact/mathutil.c's FindDistance2D()/FindDistance3D().
The main use is to allow passing dx/dy instead of sprite positions; the code
that actually uses this is not committed.
git-svn-id: https://svn.eduke32.com/eduke32@4579 1a8010ca-5511-0410-912e-c29ae57300e0
They classify former uses of int64_t into two classes:
- coord_t: those that represent coordinates and need 64-bit precision for
proper functioning with large values
- inthi_t: those that were added for e.g. casts merely to prevent undefined
behavior on overflow
DONT_BUILD.
git-svn-id: https://svn.eduke32.com/eduke32@4578 1a8010ca-5511-0410-912e-c29ae57300e0
This could happen when building outside the "classic" grid limits and would
then lead to e.g. incorrect loop assignment on sector splitting. Bug reported
by MetHy.
git-svn-id: https://svn.eduke32.com/eduke32@4572 1a8010ca-5511-0410-912e-c29ae57300e0
Inspired by
http://forums.duke4.net/topic/7506-tror-question/page__view__findpost__p__199151
the corruption checker now checks for certain conditions of the loops of each
sector. Recall that CW loops are outer and CCW loops are inner.
- If a sector has no or more than one outer loop, count that as corruption
(level 4 and 3, respectively).
- (Disabled) For sectors with exactly one outer loop, check that all inner
ones are inside it. This is currently not compiled due to an asymmetry of
loopinside() for degenerate cases, similar to pre-r3898 inside().
git-svn-id: https://svn.eduke32.com/eduke32@4569 1a8010ca-5511-0410-912e-c29ae57300e0
There are no intended changes of functionality, it's readability tweaks only.
git-svn-id: https://svn.eduke32.com/eduke32@4568 1a8010ca-5511-0410-912e-c29ae57300e0
-sdlayer: Apple and GTK are not mutually exclusive.
-Makefile.common: Specify WITHOUT_GTK?=1 by default.
-Makefile: Limit linking with the QuickTime framework to OS X 10.6 and below.
One fix from me:
-osxbuild.sh: Fix building tools.
git-svn-id: https://svn.eduke32.com/eduke32@4548 1a8010ca-5511-0410-912e-c29ae57300e0
1 compiler error.
2 compiler warnings.
1 runtime warning.
Note that at this time, CPLUSPLUS=1 LUNATIC=1 fails to build due to C++ function mangling, and Win64 Lunatic crashes.
DONT_BUILD.
git-svn-id: https://svn.eduke32.com/eduke32@4540 1a8010ca-5511-0410-912e-c29ae57300e0
1. The application must specify its proper name and technical name. Instead of eduke32_or_mapster32.crash.log, we now have eduke32.crash.log and mapster32.crash.log.
2. The exception handler will display a message box informing the user of a crash and requesting they send in the crash log. The box has three options: "Quit", the DLL's current behavior, "Continue", which passes the exception to the next handler, and "Ignore", which resumes execution immediately. These should allow the user to skip bogus exceptions picked up by ebacktrace, such as one I get with my laptop that causes EDuke32 no issues.
git-svn-id: https://svn.eduke32.com/eduke32@4538 1a8010ca-5511-0410-912e-c29ae57300e0
When that mode is enabled (see below for caveats), wall textures that have a
non-power-of-two height (call it 'h') will be modified to look like in classic:
Let 'H' be the next power of two greater than 'h'. The texture will be uploaded
with height 'H', made up from 'h' hlines of the original texture, followed by
'H'-'h' first hlines of the same.
No panning "corrections" will take place. The mode is disabled by default.
Caveats/notes:
* the mode requires that r_hightile is disabled
* it is not implemented in Polymer
* in the Lunatic build, it is ineffective when a VX map is loaded, as those
display walls with NPOT height textures correctly
git-svn-id: https://svn.eduke32.com/eduke32@4498 1a8010ca-5511-0410-912e-c29ae57300e0
... and cull code that is dead with the X*alloc() versions since they never
return NULL on requesting memory.
Use something like
git grep '[^Xx]\(m\|c\|re\)alloc *('
and
git grep '[^Xx]strdup *('
to see places where I left the B*alloc() calls intact.
BUILD_LUNATIC.
git-svn-id: https://svn.eduke32.com/eduke32@4491 1a8010ca-5511-0410-912e-c29ae57300e0
These wrap the x*alloc or xstrdup functions in compat.c. The handler gets passed
__FILE__, __LINE__ and __func__ (if available) in debugging builds.
Terminating the application process immediately in case of allocation failure
will let us prune many error handling paths and simplify a good portion of code.
git-svn-id: https://svn.eduke32.com/eduke32@4490 1a8010ca-5511-0410-912e-c29ae57300e0
- in mdloadskin() and gloadtile_hi(), use new function check_nonpow2()
(bit-twiddling) instead of loop
- Replace a couple of missed literals with CACHEAD_* enum labels
git-svn-id: https://svn.eduke32.com/eduke32@4488 1a8010ca-5511-0410-912e-c29ae57300e0
Note: in polymer.c, this was buggy (but harmless) because .r is a char:
if ((uint32_t)hictinting[MAXPALOOKUPS-1].r & 0xFFFFFF00) != 0xFFFFFF00)
git-svn-id: https://svn.eduke32.com/eduke32@4487 1a8010ca-5511-0410-912e-c29ae57300e0
Specifically, the handling of one-way walls ([1] in Mapster32) was slightly
different.
git-svn-id: https://svn.eduke32.com/eduke32@4484 1a8010ca-5511-0410-912e-c29ae57300e0
Mostly, these are rendering-related variables. We keep *writing* them to
mapster32.cfg so that older Mapster32 versions can be used side-by-side
for now.
git-svn-id: https://svn.eduke32.com/eduke32@4462 1a8010ca-5511-0410-912e-c29ae57300e0
This mode does not take a sector's visibility values into account for uniform
darkening of a texture. It is more faithful to classic with respect to
fog/visibility *distances*, but may be less so as far as the "tint" of the
texture due to visibility attenuation is concerned.
git-svn-id: https://svn.eduke32.com/eduke32@4459 1a8010ca-5511-0410-912e-c29ae57300e0
That is, for every for palette (now, even ones created by the user) the last
shade is the (or rather, a color close to the) full fog color by construction.
In loadpalette(), reject a PALETTE.DAT with only one shade level.
git-svn-id: https://svn.eduke32.com/eduke32@4458 1a8010ca-5511-0410-912e-c29ae57300e0
Concretely, for all 1 <= i <= 255, make graypalookup[31][i] := graypalookup[31][0].
Also document 'fogpal' in the Lunatic doc.
git-svn-id: https://svn.eduke32.com/eduke32@4456 1a8010ca-5511-0410-912e-c29ae57300e0
Also,
- add a static assertion to engine.c
- gameexec.c: compile out an unused function in Lunatic builds
git-svn-id: https://svn.eduke32.com/eduke32@4451 1a8010ca-5511-0410-912e-c29ae57300e0
This makes the game not process as many ticks as have elapsed during the saving
afterwards.
git-svn-id: https://svn.eduke32.com/eduke32@4450 1a8010ca-5511-0410-912e-c29ae57300e0
The extended PALETTE.DAT format can now carry a trailing byte, called
'lognumalphatables'. If it is present, the engine assumes that alpha blending
tables are loaded at blend indices 1 to 1<<lognumalphatables and handles
rotatesprite's alpha and spriteext[].alpha accordingly.
The ability to save this new byte is added to engine.lua:savePaletteDat()
and the corresponding convenience functions in shadexfog.lua.
Examples test/screentext.con and test/sprite_access.con are updated for
demonstration purposes.
git-svn-id: https://svn.eduke32.com/eduke32@4426 1a8010ca-5511-0410-912e-c29ae57300e0
The overall situation is thus as follows:
- in C and from Lua, that member is called 'fogpal'
- In CON (both C-CON and LunaCON), it's 'fogpal' with 'alignto' being an alias;
'filler' wasn't available before, either.
- In m32script, it's 'fogpal' or 'alignto' ('filler' is not available any more)
git-svn-id: https://svn.eduke32.com/eduke32@4416 1a8010ca-5511-0410-912e-c29ae57300e0
That is, if that member has a nonzero value, that one is taken for the color of
the GL fog instead of (most of the time) sector[].floorpal. "Sky sectors", that
is, skyboxes or simply parallaxed floors/ceilings are *not* handled.
git-svn-id: https://svn.eduke32.com/eduke32@4415 1a8010ca-5511-0410-912e-c29ae57300e0
Also, extend the documentation for r_usenewshading a little. Specifically,
note when the different modes were introduced. BUILD_LUNATIC.
git-svn-id: https://svn.eduke32.com/eduke32@4410 1a8010ca-5511-0410-912e-c29ae57300e0
It was plainly noticeable in the editor's tile info display, for
example.
git-svn-id: https://svn.eduke32.com/eduke32@4409 1a8010ca-5511-0410-912e-c29ae57300e0
This is done by factoring out some polymer.c:polymer_getbuildmaterial() code
into polymer_eligible_for_artmapping() and using that. The one in polymer.c
can still determine to not use ART mapping on some ("rare"?) occasions, so
it's not 100% in sync.
git-svn-id: https://svn.eduke32.com/eduke32@4407 1a8010ca-5511-0410-912e-c29ae57300e0
- factor out eligible_for_tileshades()
- remove some rendmode 1 code
- move some auto vars further down
git-svn-id: https://svn.eduke32.com/eduke32@4406 1a8010ca-5511-0410-912e-c29ae57300e0
With this mode, the near distance of the linear fog is always 0, unlike being
negative with r_usenewshading 2.
git-svn-id: https://svn.eduke32.com/eduke32@4393 1a8010ca-5511-0410-912e-c29ae57300e0
Make Mapster32 save as map-text if there is a wall with non-zero 'blend'.
git-svn-id: https://svn.eduke32.com/eduke32@4368 1a8010ca-5511-0410-912e-c29ae57300e0
For a 1680x1050 scene standing directly in front of a mirror, this increases
FPS from ~90 to ~95 for me.
git-svn-id: https://svn.eduke32.com/eduke32@4363 1a8010ca-5511-0410-912e-c29ae57300e0
One potential use is to add r_togglecomposition "0" if you don't like Aero turning on and off every time you switch between 2D mode and 32-bit 3D mode.
git-svn-id: https://svn.eduke32.com/eduke32@4344 1a8010ca-5511-0410-912e-c29ae57300e0
- Mapster32: before loading LOOKUP.DAT, set palookup[0][239]=239 to
make an identity map of the base shade table's shade 0
- Rewrite color index remapping case of makepalookup() for clarity
BUILD_LUNATIC.
git-svn-id: https://svn.eduke32.com/eduke32@4336 1a8010ca-5511-0410-912e-c29ae57300e0
As opposed to the previous way, where the first fog pals was <numlookups>+1,
where <numlookups> is the first byte value of LOOKUP.DAT. This allows to
pack e.g. lookups [1 .. 25] and [30 .. <lastpal>] into LOOKUP.DAT and have fog
pals be generated at pals [26 .. 29] (i.e. the additional lookups don't
shift the fog pals, making user maps depending on these numbers not look as
intended.)
git-svn-id: https://svn.eduke32.com/eduke32@4335 1a8010ca-5511-0410-912e-c29ae57300e0
Also, error if didn't read enough data and account for TITLE and REALMS
swap between basepal number and on-disk order (sigh).
git-svn-id: https://svn.eduke32.com/eduke32@4334 1a8010ca-5511-0410-912e-c29ae57300e0
Lunatic: also, add an additional argument 'moreblends' to
engine.savePaletteDat() and document that function.
See test/shadexfog.lua for a "user-friendly" wrapper shadexfog.save() which
prints success or errors.
git-svn-id: https://svn.eduke32.com/eduke32@4312 1a8010ca-5511-0410-912e-c29ae57300e0
On the C side, slightly rewrite (now) getclosestcol_lim() for clarity.
git-svn-id: https://svn.eduke32.com/eduke32@4308 1a8010ca-5511-0410-912e-c29ae57300e0
Also, fix a missed .filler -> g_ambiencePlaying[] rewrite in the editor.
In C-CON, 'detail' is kept as alias to C-side (former) .filler / now .blend,
but that name is *deprecated*.
git-svn-id: https://svn.eduke32.com/eduke32@4305 1a8010ca-5511-0410-912e-c29ae57300e0
This fixes updating the viewscreen in Polymer's ART mapping mode.
git-svn-id: https://svn.eduke32.com/eduke32@4304 1a8010ca-5511-0410-912e-c29ae57300e0
Currently, sprite[].filler (to be renamed) selects a blending table. If none
is installed with that index, the blending table at index 0 is taken (which is
read from PALETTE.DAT on engine startup). Thus, for now, the non-Lunatic build
will never use a custom blending table because there's no way to register one!
But the functionality is there.
git-svn-id: https://svn.eduke32.com/eduke32@4301 1a8010ca-5511-0410-912e-c29ae57300e0
- use SPRITES_OF* iterator macros
- new function: E_SpriteIsValid(); add some safety checks
(e.g. hitag used as sprite index)
- the rest is mostly "the usual code transformations"
git-svn-id: https://svn.eduke32.com/eduke32@4281 1a8010ca-5511-0410-912e-c29ae57300e0
In the non-Lunatic build naturally, since the Lunatic one uses its own member.
git-svn-id: https://svn.eduke32.com/eduke32@4275 1a8010ca-5511-0410-912e-c29ae57300e0
EDuke32 has me so accustomed to straight C... using a template and pass-by-reference makes my day.
git-svn-id: https://svn.eduke32.com/eduke32@4267 1a8010ca-5511-0410-912e-c29ae57300e0
In simple cases, it should be added. Combinations of alignment across TROR
boundaries and bottom-swapping will probably not work right.
git-svn-id: https://svn.eduke32.com/eduke32@4265 1a8010ca-5511-0410-912e-c29ae57300e0
Concretely, in the "behind, outside" case, draw the sprite early if any of its
2 or 4 corner points are behind the masked wall, and inside the cone or on the
other outside.
git-svn-id: https://svn.eduke32.com/eduke32@4264 1a8010ca-5511-0410-912e-c29ae57300e0
This prevents a crash and maybe other badnesses when e.g. loading a map
with ART tiles, doing something that allocates a reserved tile (such as
looking into a viewscreen), then loading another map where the tilesiz*
of TILE_VIEWSCR is restored to 0.
git-svn-id: https://svn.eduke32.com/eduke32@4259 1a8010ca-5511-0410-912e-c29ae57300e0
When a map named <filename>.map is loaded (<filename> may also contain directory
separators), the engine checks for existence of <filename>_XX.art in the virtual
file system, where XX is a 0-padded number from 00 to 19. It loads a consecutive
sequence of these ART files, i.e. aborts whenever a number in the sequence isn't
found (in contrast to normal ART loading).
Restrictions:
- the per-map ART files must not reside in ZIP files
- if a tile number is attempted to be overridden that has a dummytile or is
cache1d-locked, per-map ART loading fails
On failure, the map is still loaded, but a diagnostic message is output to the
log/OSD.
Loaded per-map ART data are cleared whenever the map is "left". In particular:
- whenever another map is loaded
- in the editor: when a new map is started
- in the game: after the bonus ending screen of a finished level, after going
to the title screen via the menu
A final note: file names are supposed to be looked up and compared
case-sensitively. That is, <filename> must match EXACTLY between the map's and
per-map ART one; 'art' must be lowercase. Otherwise, the cookie monster will
come and eat you!
BUILD_LUNATIC.
git-svn-id: https://svn.eduke32.com/eduke32@4257 1a8010ca-5511-0410-912e-c29ae57300e0
Available in the game and editor. Provide test/shadexfog.lua, containing a
function to creating a set of 32 shade tables corresponding to different shades
of the same fog palookup table, together with some tests and convenience
functions.
Also,
- Add gv.LUNATIC_CLIENT{,_EDUKE32,_MAPSTER32}
- Add LUNATIC_FIRST_TIME in the global env for the game
- defs_m32.lua: add reload() convenience function
- Failed attempt at recreating the base shade table. It is NOT a linear ramp
of the base palette colors to (0,0,0). That is, it's not created by
build/util/transpal.exe!
git-svn-id: https://svn.eduke32.com/eduke32@4236 1a8010ca-5511-0410-912e-c29ae57300e0
The scrollwheel is unique among PC input because it has no innate "hold length". Previously, the layers gave the mousewheel a fake hold length to allow the not-necessarily-synchronous game/editor code to pick up the input before the layers marked it as "no longer pressed". This passed under Windows, but it didn't slide under SDL.
Besides the two problems listed above, it also potentially limited the rate of weapon selection, where scrolling too fast would not register every clicks. [Unrelatedly, this is still the case when you scroll faster than the game's own tickrate, but addressing that would require rewriting input handling to go through a list of "events" for each tic instead of looking at overall pressed/unpressed states.]
git-svn-id: https://svn.eduke32.com/eduke32@4200 1a8010ca-5511-0410-912e-c29ae57300e0
This requires enabling the same workaround as for Linux. Maybe it's not
system-dependent at all.
git-svn-id: https://svn.eduke32.com/eduke32@4193 1a8010ca-5511-0410-912e-c29ae57300e0
I was checking for GCC >= 4.7, while Clangs pretends to be GCC 4.2. Use a
feature test macro instead. The comment I made in r4161 regarding GCC vs.
Clang code was wrong. Now, Clang generates slightly faster code for these cases
(solid and masked 4-pixel wide vlines).
git-svn-id: https://svn.eduke32.com/eduke32@4182 1a8010ca-5511-0410-912e-c29ae57300e0
Which was especially noticeable in the Lunatic build. Also, use
glPushAttrib/glPopAttrib there.
git-svn-id: https://svn.eduke32.com/eduke32@4176 1a8010ca-5511-0410-912e-c29ae57300e0
This "simply" requires an appropriate factor in the projection matrix.
Also, get rid of a special-case factor for >=1.6 aspects (making HUD models
wider then, I think).
git-svn-id: https://svn.eduke32.com/eduke32@4172 1a8010ca-5511-0410-912e-c29ae57300e0
Known as getactor/sprite[].xpanning from CON, values are 0-255. It should be
assumed that this is only valid for wall-aligned sprites (currently, face
sprites also pan).
A test is provided in lunatic/test/animatesprites.con. It should be run in E1L2
and only with the Lunatic build.
git-svn-id: https://svn.eduke32.com/eduke32@4170 1a8010ca-5511-0410-912e-c29ae57300e0
By detecting them in the SDL_KEYDOWN events. This list is getting a bit
ridiculous...
git-svn-id: https://svn.eduke32.com/eduke32@4163 1a8010ca-5511-0410-912e-c29ae57300e0
Compiling a 32-bit NOASM build resulted in code containing a MOVAPS instruction
that accessed a memory location not aligned to 16 bytes (MinGW, GCC 4.8).
git-svn-id: https://svn.eduke32.com/eduke32@4162 1a8010ca-5511-0410-912e-c29ae57300e0
The functions mvlineasm1, mvlineasm4 and tvlineasm2 can now be set to clamp
the vertical texture coordinate (vplc), preventing the unsightly stray lines
on the bottom of non-y-flipped sprites. (The first part of this effort, r3483,
handled their top).
However, this is only enabled for the mvlineasm ones: the vectorized variants
suffered almost no slowdown (even though a PADDUSD SSE instruction would be a
nice thing to have), while it was pretty significant for the sequential
translucent ones.
Summarizing, this leaves two cases where stray lines may appear in the non-ASM
build (the saturation is NYI for a.nasm):
- at the bottom of y-flipped sprites
- at the bottom of translucent sprites (can be toggled by #define)
Another observation is that recent GCC generates much faster code for this
stuff than Clang from SVN.
git-svn-id: https://svn.eduke32.com/eduke32@4161 1a8010ca-5511-0410-912e-c29ae57300e0
For a full 1680x1050 screen drawing a solid/masked wall, the FPS increases
from 118 to 133 and from 114 to 116 (respectively) for me.
Guarded by the macro USE_VECTOR_EXT in the source.
git-svn-id: https://svn.eduke32.com/eduke32@4160 1a8010ca-5511-0410-912e-c29ae57300e0
This fixes some init and update anomalies: before,
- init would always be the graphical mode (0), but writing
merely "osdtextmode" would switch it
- osdtextmode would not get saved to the cfg
git-svn-id: https://svn.eduke32.com/eduke32@4139 1a8010ca-5511-0410-912e-c29ae57300e0
These were too much "looking into the implementation of a module" for my
taste, hence OSD_GetShadePal() earlier.
git-svn-id: https://svn.eduke32.com/eduke32@4138 1a8010ca-5511-0410-912e-c29ae57300e0
This renders tweaks like r4122 unnecessary.
Also, note a "bug" with the program generating that C code.
git-svn-id: https://svn.eduke32.com/eduke32@4135 1a8010ca-5511-0410-912e-c29ae57300e0
So that there are no error messages like "M32 file `a.m32.m32' not found."
Also, in build.c's 'save as' code, add a bound check that would probably
always pass in practice, but looks a bit safer and may fail in very cornerly
cases.
git-svn-id: https://svn.eduke32.com/eduke32@4124 1a8010ca-5511-0410-912e-c29ae57300e0
Instead of directly in build.c:app_main(). Previously, it was possible that an
argument to an option, e.g. "WGR2" with "-game_dir WGR2", would get interpreted
as the initial map to load. This in turn would attempt to execute WGR2.cfg as
sequence of OSD commands -- where that file is an EDuke32 config file instead!
Now: maps are passed without any options, e.g. "mapster32 [opts...] debug.map".
The map file name is not added to the "additional parameters" for map testing.
git-svn-id: https://svn.eduke32.com/eduke32@4121 1a8010ca-5511-0410-912e-c29ae57300e0
Because the one implemented on top of X shows a special mark for tabs instead of
whitespace.
Also, in wm_msgbox(), use vsnprintf.
git-svn-id: https://svn.eduke32.com/eduke32@4120 1a8010ca-5511-0410-912e-c29ae57300e0
This fixes ESC in certain situations such as the editor map selection menu.
git-svn-id: https://svn.eduke32.com/eduke32@4117 1a8010ca-5511-0410-912e-c29ae57300e0
