That is, "clang -ftrapv" builds don't abort almost immediately after entering
a level.
There are various classes of overflow bugs, needing different handling:
- Some texture mapping code was written with signed integer wrapping semantics
in mind. In some places, we're able to get away with unsigned casts.
- sometimes, we really need a wider range, like when calculating distances or
dot products
- negating INT_MIN. Here, we cast to int64_t temporarily. Note that if the
result is 32-bit wide, no 64-bit code may actually need to be generated.
- shifting into a signed integer's sign bit. We cast to uint32 here.
- in hitscan(), at the "abyss crash prevention code" comment, it's clearly
the other code that is better...
This is not merely done for pedantry, but rather makes it easier to track down
overflow bugs having a real impact on the game.
git-svn-id: https://svn.eduke32.com/eduke32@2784 1a8010ca-5511-0410-912e-c29ae57300e0
It's time to replace some int32s with 64-bit ints in core engine functions.
The problem is that for example, the dot product is taken between vectors that
may be the difference between two arbitrary points in a sector, so even if one
sticks to the "no blue walls" rule, that doesn't guarantee freedom from
overflows.
git-svn-id: https://svn.eduke32.com/eduke32@2761 1a8010ca-5511-0410-912e-c29ae57300e0
Bang! One more invalid internals exposure squashed. Maybe one day we'll be able
to call EDuke32 "stable"...
Actually the offending sprite IS drawn as face sprite with shade 32 and xrepeat
and yrepeat 255 for the convenience of the CON coder who will have to debug it.
git-svn-id: https://svn.eduke32.com/eduke32@2744 1a8010ca-5511-0410-912e-c29ae57300e0
Those were happening when wall drawing in an upper/lower layer "closed" too
much of the "curtain" (umost/dmost) for subsequent drawing in that portal.
Example: looking down the sewer manhole at the beginning of Retaliation will
now render without HOM.
git-svn-id: https://svn.eduke32.com/eduke32@2739 1a8010ca-5511-0410-912e-c29ae57300e0
Also, correct the loadboard() return value checks in premap.c to be
aware of a returned -2 ("wrong map version").
git-svn-id: https://svn.eduke32.com/eduke32@2694 1a8010ca-5511-0410-912e-c29ae57300e0
This should be more correct, since previously, the saved PNGs were way too
contrasty when looked from within a browser. Now, they'll always be
slightly darker than in-game (assuming one has a gamma greater than 1).
git-svn-id: https://svn.eduke32.com/eduke32@2673 1a8010ca-5511-0410-912e-c29ae57300e0
- Make MAXXDIM and MAXYDIM macros be 860 and 490, respectively. tueidj says
max. screen size is 848x480, but there are oob access bugs in the classic
renderer when running with the maximum possible resolution
- don't allocate additional sector/wall storage for Mapster (saves ~16k)
- Add ud.config.UseJoystick = 1 in config.c, but make it conditional on Wii
- remove one comment I no longer believe true, add one TODO comment
git-svn-id: https://svn.eduke32.com/eduke32@2645 1a8010ca-5511-0410-912e-c29ae57300e0
Previously, we checked whether the faded palette has changed (by way of CRC)
and invalidated the textures then unless the preserve flags were set. This
however could lead to wrongly invalidating them under unfortunate circumstances,
e.g. basepal change from CON + tints at the same time before r2620 which
reverted r2232.
Now, only invalidate them if the corresponding preserve flags are clear AND
* the base palette has really changed OR
* the palette CRC changed and we were running on software gamma
The latter means that performance-killing invalidations may still happen on
GL platforms lacking HW gamma (for ATI, it's currently only disabled in 8-bit
fullscreen).
Also have a new global 'basepalreset' to fake a basepal change for
setbrightness(), currently used when changing renderers so that going from
Polymer to Polymost and back again will invalidate the Polymer textures
on the second change, potentially re-applying a basepal highpal. (Still
with me?)
git-svn-id: https://svn.eduke32.com/eduke32@2636 1a8010ca-5511-0410-912e-c29ae57300e0
Because gltexinvalidate*() could be called too often when setgamepalette is used
while having a tint overlaid else. Pending thorough analysis/reworking of the
32-bit mode base palette handling / texture invalidation.
git-svn-id: https://svn.eduke32.com/eduke32@2620 1a8010ca-5511-0410-912e-c29ae57300e0
This has been there since searchbottomwall introduction in r1466.
git-svn-id: https://svn.eduke32.com/eduke32@2604 1a8010ca-5511-0410-912e-c29ae57300e0
Previously, we only set the viewingrange according to the physical screen's
dimensions, but didn't correct yxaspect for a potential non-square pixel
ratio.
git-svn-id: https://svn.eduke32.com/eduke32@2603 1a8010ca-5511-0410-912e-c29ae57300e0
- factor out many identical checks in a convenient function; some messages
may read slightly differently now and tile ranges may be handled more strictly
(error out if one of the limits is invalid)
- factor out two instances of identical (up to one arg) code into
tile_from_truecolpic
- factor out setting picsiz[] and stuff into set_picsizanm
- some checks
- Make "undefmodelof" non-functional and warn.
- in "animtilerange", if the tile difference is >= 64, error out since we
can't store it in picanm[]
git-svn-id: https://svn.eduke32.com/eduke32@2588 1a8010ca-5511-0410-912e-c29ae57300e0
This requires one tweak in drawrooms' umost/dmost setup to prevent oob access.
Specifically, a coordinate difference of 0 is allowed. In the classic renderer,
this would mean a one-pixel (real screen coords) height or width. In Polymost,
it would currently mean a one-pixel height and zero-pixel width, but this might
be subject to change.
git-svn-id: https://svn.eduke32.com/eduke32@2574 1a8010ca-5511-0410-912e-c29ae57300e0
(That is, the base shade table.) Before, we allocated each palookup buffer.
For a vanilla setup, this means that we're now saving 224*32*256 ~= 1.8 megs,
which might be interesting for low-memory gadgets.
git-svn-id: https://svn.eduke32.com/eduke32@2571 1a8010ca-5511-0410-912e-c29ae57300e0
First, it's unlikely in our day and age. Second, they're always free'd at the end,
so allocache'ing them is incorrect.
git-svn-id: https://svn.eduke32.com/eduke32@2570 1a8010ca-5511-0410-912e-c29ae57300e0
Also,
- use this in game.c and astub.c palookup loading code
- when makepalookup() is passed a 0 palnum, return early. This means that
'fogpal' will silently fail when attempting to change pal 0.
- in 'makepalookup' DEF command, error out if passed a pal of 0.
git-svn-id: https://svn.eduke32.com/eduke32@2569 1a8010ca-5511-0410-912e-c29ae57300e0
Alongside, these make into into the header:
- the 'tokenlist' type (a typedef'd struct)
- the T_EOF and T_ERROR enumeration values
git-svn-id: https://svn.eduke32.com/eduke32@2549 1a8010ca-5511-0410-912e-c29ae57300e0
- Help window text cleaned and made more consistent between game and editor
- Added help entry for "-clipmap"
- Log text for using CON, DEF, and RTS files has been made consistent
- All instances of '%s' have been replaced with \"%s\" because ' is a valid filename character. (At least on Windows.)
git-svn-id: https://svn.eduke32.com/eduke32@2538 1a8010ca-5511-0410-912e-c29ae57300e0
There are instances where oob picnums may propagate to that function, so
protect it. The digitanumber[z] bound check is actually made more permissive,
but could also just as well be removed now.
git-svn-id: https://svn.eduke32.com/eduke32@2533 1a8010ca-5511-0410-912e-c29ae57300e0
This makes the differences in these codes stand out much more clearly.
git-svn-id: https://svn.eduke32.com/eduke32@2526 1a8010ca-5511-0410-912e-c29ae57300e0
We shouldn't assume a particular bytes-per-line value and use ylookup[] instead.
Specifically, windowed modes on Windows use a frame buffer that always has odd
x dimension.
git-svn-id: https://svn.eduke32.com/eduke32@2505 1a8010ca-5511-0410-912e-c29ae57300e0
- in Mapster, pre-form the default 10 clip map names before returning from
G_CheckCommandLine() so it gets loaded even if we passed no cmdline args.
- malloc + strlen + strcpy --> strdup
- don't need to spank dead variables ;)
- we may call calloc with zero size, which isn't bad by itself, but asserting
for non-null afterwards is. Allocs of 0 are implementation-defined, and may
well return a null pointer (C99 7.20.3).
git-svn-id: https://svn.eduke32.com/eduke32@2502 1a8010ca-5511-0410-912e-c29ae57300e0
Gives a couple more fps for scenes where much screen estate is covered by
stuff, like when holding the devastator.
git-svn-id: https://svn.eduke32.com/eduke32@2499 1a8010ca-5511-0410-912e-c29ae57300e0
With the same setup as before, a screen-filling translucent wall (with nothing
drawn behind it) renders at about 7 fps faster (from 60-something fps initially)
git-svn-id: https://svn.eduke32.com/eduke32@2498 1a8010ca-5511-0410-912e-c29ae57300e0
These two functions draw a vertical line 4 neighboring pixels at a time.
This gives a significant speed boost for a full screen solid and masked wall
scene for x86_64 (where we have plenty of registers), about 60 --> 76 fps.
git-svn-id: https://svn.eduke32.com/eduke32@2497 1a8010ca-5511-0410-912e-c29ae57300e0
Also, a very minor change in the con/def module code. (int --> int32_t)
git-svn-id: https://svn.eduke32.com/eduke32@2495 1a8010ca-5511-0410-912e-c29ae57300e0
Previously, if the sprite turned really out to be in void space, either freelist
inconsistency (before the list rewrite) or oob access (now) would happen.
Also add an bound-checking assert() for insertsprite's sectnum argument (it's not
a bound check!)
git-svn-id: https://svn.eduke32.com/eduke32@2480 1a8010ca-5511-0410-912e-c29ae57300e0
New engine variable 'int32_t Numsprites', not yet saved into savegames
or mapstates. (The capitalization is to distinguish it from the often-used
'numsprites' locals or structure member names.
In the editor, get rid of updatenumsprites().
git-svn-id: https://svn.eduke32.com/eduke32@2478 1a8010ca-5511-0410-912e-c29ae57300e0
Previously, the lists starting at headspritestat[MAXSTATUS] and
headspritesect[MAXSECTORS] were both used as sprite freelists and were always
in complete synchrony. Now, make only the statnum list keep the free sprites.
This way, it has no CON compatibility implications because
headspritesect[MAXSECTORS] is inaccessible there. Leave the array at its
original size for now.
git-svn-id: https://svn.eduke32.com/eduke32@2477 1a8010ca-5511-0410-912e-c29ae57300e0
The major outside-visible change is that this fixes the sound cutoff bugs that
happened because newly-spawned sprites took the place of those whose sounds
had not yet finished playing.
Besides, there are these changes:
- remove deletesprite{sect,stat}
- we have a new engine variable 'tailspritefree' that keeps track of the
sprite freelist tail
- we need to store it in savegames and mapstates, so bump the savegame
minor version
git-svn-id: https://svn.eduke32.com/eduke32@2470 1a8010ca-5511-0410-912e-c29ae57300e0
This is simply done by using the two list helper functions defined earlier
instead of a deletespriteX/insertspriteX pair.
git-svn-id: https://svn.eduke32.com/eduke32@2469 1a8010ca-5511-0410-912e-c29ae57300e0
Factor out two helper functions, do_deletesprite{sect,stat}
and do_insertsprite_at_headof{sect,stat}.
git-svn-id: https://svn.eduke32.com/eduke32@2466 1a8010ca-5511-0410-912e-c29ae57300e0
These operations aren't executed THAT often to warrant exposing four internal
functions. Besides, I'll be hacking on them and need everything laid out
cleanly. Should somebody complain, they can go change them to __fastcall or
something.
git-svn-id: https://svn.eduke32.com/eduke32@2465 1a8010ca-5511-0410-912e-c29ae57300e0
In changesprite{stat,sect}, don't check for deletesprite{stat,sect} failure,
because the only way it can fail was already checked for above. So, this is
essentially dead code elimination.
git-svn-id: https://svn.eduke32.com/eduke32@2462 1a8010ca-5511-0410-912e-c29ae57300e0
- forgot a glogy --> logy in a-c.c
- comment out stretchhline and slopevlin2 in a.nasm, the former also in a-c.c
- make transmaskvline2 use a uintptr_t where appropriate
git-svn-id: https://svn.eduke32.com/eduke32@2448 1a8010ca-5511-0410-912e-c29ae57300e0
When drawing from inside the mirror, there may be no 'bunches' (BUILD lingo,
not TROR ones) collected, but following code used the first one unconditionally.
git-svn-id: https://svn.eduke32.com/eduke32@2427 1a8010ca-5511-0410-912e-c29ae57300e0
The apparent horizon (in units proportional to pixels by a constant factor)
was being calculated without normalizing to viewingrange, resulting a
discrepancy between the drawn and actual up/down angle with non-default
viewingranges (e.g. with r_usenewaspect=1, or when setting it from setaspect).
This also means that the center-of-aim will now be the same regardless of
the aspect.
This fixes many derived bugs:
* inconsistency w/ crosshair when shooting a gun and aiming up/down (obviously)
* bugs resulting from being able to look up/down too far (garbage non-slope
texture-mapping and viewing BIGORBIT skies above the borders in classic).
Note: mods that had workaround logic for this bug will have to remove it...
git-svn-id: https://svn.eduke32.com/eduke32@2414 1a8010ca-5511-0410-912e-c29ae57300e0
This has no measureable difference. Oh well, maybe it benefits weaker
machines, since the generated loop code should definitely be prettier.
git-svn-id: https://svn.eduke32.com/eduke32@2408 1a8010ca-5511-0410-912e-c29ae57300e0
This doesn't give much, maybe 0.5 fps more, probably because the
loop body is pretty long.
git-svn-id: https://svn.eduke32.com/eduke32@2406 1a8010ca-5511-0410-912e-c29ae57300e0
(Classic renderer). We don't need 64-bit integers; also use the nice
asm version of divscale32() if possible.
git-svn-id: https://svn.eduke32.com/eduke32@2390 1a8010ca-5511-0410-912e-c29ae57300e0
This was discovered because viewscreens (TILE_VIEWSCR) ceased being drawn
with the MAXTILES increase (reported by Hank, who also did the bisect).
git-svn-id: https://svn.eduke32.com/eduke32@2376 1a8010ca-5511-0410-912e-c29ae57300e0
That is, everyplace a user input is to be converted. The only remaining
instance of atoi() is now atoi(s_buildDate).
git-svn-id: https://svn.eduke32.com/eduke32@2374 1a8010ca-5511-0410-912e-c29ae57300e0
Internally, the last argument to neartag is now a pointer to a function
int32_t (*blacklist_sprite_func)(int32_t i),
which is supposed to return 1 if sprite[i] should NOT be considered for hitting.
This is now used in the hard-coded neartag() calls in sector.c, but not in any
way in CON (there's neither a C blacklist function provided, nor is there a
possibility to define one in CON). There, all sprites with picnums >=1 and <=10
(i.e. the effectors) will be blacklisted. This remedies problems where such
sprites would get in the way of switches.
Note that a whitelist approach (only consider a predefined set, namely those
picnums which will be checked afterwards) has back-compatibility implications
since people may have used e.g. lotagged window sprites to cover a switch.
Also, the >=1 to <=10 range is [sic] (the static, not dynamic values are used),
since anyone redefining effector picnums is clearly out of their mind.
git-svn-id: https://svn.eduke32.com/eduke32@2373 1a8010ca-5511-0410-912e-c29ae57300e0
The bulk are whitespace changes; the only exception is a
"if (expr) { stmt; }" --> "if (expr) stmt;"
git-svn-id: https://svn.eduke32.com/eduke32@2353 1a8010ca-5511-0410-912e-c29ae57300e0
The reason was that drawmasks(), which was where the resetting of the fake
RoR textures happened under Polymer, can be called from polymer_displayrooms()
when mirrors are present. I was assuming that drawmasks() was only ever called
after the 'main' drawrooms and that hence the sequence of fake RoR tile tweaking
was set, reset, set, reset, ... which it was not.
Now, do the resetting explicitly after every drawmasks() in the editor.
git-svn-id: https://svn.eduke32.com/eduke32@2350 1a8010ca-5511-0410-912e-c29ae57300e0
I find this cleaner for two reasons. First, lotags won't get in the way of
neartag(), so that doors and switches near constrained TROR walls can be
pushed. Second, this removes the information leak so that CON coders aren't
tempted to read that values directly before a proper API for that comes along.
When writing out a map in-game (e.g. with dndebug), restore the respective
members so that it can be loaded in the editor again.
git-svn-id: https://svn.eduke32.com/eduke32@2349 1a8010ca-5511-0410-912e-c29ae57300e0
getzrange() returns the floor rather than the sprite if their heights are
equal. Now, make an exception for extended floors (analogously, ceilings).
git-svn-id: https://svn.eduke32.com/eduke32@2347 1a8010ca-5511-0410-912e-c29ae57300e0
- guard cansee() against negative sectnums (return 0 immediately), since
it's often called with the player sectnum as one argument
- in resetpspritevars(), don't inset APLAYER sprite if sectnum < 0, which
can happen if a map is started in void space accidentally (e.g. from
the editor)
- two checks before accessing sector[] with a player sectnum
git-svn-id: https://svn.eduke32.com/eduke32@2342 1a8010ca-5511-0410-912e-c29ae57300e0
Internally, there's a new function fade_screen_black() that implements the
fading more efficiently than calling rotatesprite for classic, and uses
tile 0 (assumed to have at least 64x32 size) in the GL modes.
The black background of the user map selector is still done with rotatesprite
and tile BLANK in every renderer.
git-svn-id: https://svn.eduke32.com/eduke32@2312 1a8010ca-5511-0410-912e-c29ae57300e0
It looks like GCC doesn't honor the 'used' attribute with LTO. This is
"fixed" by declaring the variables in question with external linkage
(i.e. removing the 'static'). I consider such behavior a GCC bug, though.
git-svn-id: https://svn.eduke32.com/eduke32@2286 1a8010ca-5511-0410-912e-c29ae57300e0
