What changed in EDuke32, build by build

The leak happened because a struct was hashed that had uninitialized bytes in padding inserted by the compiler. The hash string in now constructed as concatenation of three CRC32s as 8-byte hex strings, i.e. the individual CRC32s are padded with leading zeros. Note to users: because of the hash change, it's sensible to delete the 'textures' and 'textures.cache' files.

References: - r3742 - r4038 (fixing savegame screenshots, but introducing the now fixed bug)

That change confounded character-based input with e.g. HUD control under certain circumstances, see http://forums.duke4.net/topic/6845-eduke32-savegame-annoyances/ Also, BUILD_LUNATIC ... just for the hit.sect -> hit.sector change.

This also exposes an actual issue. The sound definitions may be dynamically remapped, in which case they are not necessarily <=255 any more. BUILD_LUNATIC.

Most often, this had happened when casting comparison functions for qsort() like these: "int yax_cmpbunches(const int16_t *b1, const int16_t *b2)" to a function pointer type expecting "const void *". Alas, this is undefined behavior: see http://blog.frama-c.com/index.php?post/2013/08/24/Function-pointers-in-C and posts linked from it. At least two cases have not been fixed: - The savegame system maintains pointers which are either to data or a function in the generic "void *". This ought to be made into a union. - Probably also: #define OSD_ALIAS (int32_t (*)(const osdfuncparm_t*))0x1337

- Currently: only tvlineasm1 and tvlineasm2, but incomplete (no reverse translucency, nonpow2 textures will crash) - For System V AMD64 calling conventions; requires YASM

This reverts commit c7e51147f1869d42fc8365f748bb8f6bd4c6fc2f. [r3835] Conflicts: eduke32/build/src/defs.c [Dunno what caused this, but I cleaned this out. Also, the change from build.c was not reverted.]

This reduces the proportion of time the trig calculations take of the whole preprocessing from 50% to about 13%.

Also, add hitscan test to test.lua -- a crosshair-like sprite is spawned and continuously updated to the position of where the player aims at.

The offsets (that will translate the axes for .pitch/.roll rotation) were calculated improperly with screen sizes other than 1280 x 1024 and were dependent on other related variables. The fix proceeded empirically, i.e. mostly trial, error, and some educated guesswork. (It's telling that the magic constant 1280 is needed.) A test for Lunatic is added to test.lua that will rotate SEENINES. Bug pointed out by Mblackwell.

The 'colenemy' module searches for colored enemies in maps, excluding pal-21 liztroops. DONT_BUILD.

Specifically, for the tables returned with require"DEFS" etc. This allows to check if maybe parts of a bundle of files is missing (because comparisons like "spr.picnum == D.NONEXISTENT_NAME" do not yield errors normally).

Permitting to enable various debugging options. See "eduke32 -debughelp" for which ones those are: they were previously settable from defs_common.lua, and a new option 'strict' has been added that makes accesses to void sprites error. (That is, already "sprite[i]", not "sprite[i].some_member".)

- engine.c: comment out some unused decls with RELEASE=0 DEBUGANYWAY=1 - Lunatic: update dump.lua and v.lua - Lunatic: test/test_geom.lua: note a perf regression (wrt older LuaJIT, or I mis-configured / built something wrong)

That is, here: http://lunatic.eduke32.com/ or more specifically in lunatic.html. BUILD_LUNATIC.

(Auto-insertion of an inner sector from a loop happens either because the loop is created in an extended sector, or because it has been punched through a ceiling/floor into the upper/lower neighbor sector.)

- add gv.GET, gv.WEAPON, gv.RETURN - specify that sprite[].x/y/z and wall[].x/y are 32-bit ints - add an example for con.longjmp() - add an appendix listing the RETURN usage of various events

Note: the wall line will not light up periodically, but point insertion will be possible nontheless. A bit inconsistent, but I didn't want to do a full analysis of which side effects allowing a sprite highlight and a line highlight to coexist at the same time would have.

BUILD_LUNATIC, third attempt.

This makes CLANG=1 RELEASE=0 DEBUGANYWAY=2 OPTLEVEL=1 builds definitely pretty playable (only Polymost tested) while having a decent amount of sanitization.

Redundant for the same reason as the preceding one: the pointer has already been dereferenced at the point the check is made. Also, all 5 calls of that function pass a valid pointer (provided the pointers from which the expressions are derived are valid). Reported by Stack, too.

The check is dead because it is made after the fact (oob access), but the cheat string matching logic below actually assures that cheatbuflen < sizeof(cheatbuf) at all times. Exposed using the Stack tool from http://css.csail.mit.edu/stack/ . Also, in gamedef.c's definecheat handling, print the string length (19), not the buffer length (20) if the cheat was truncated.

Clang's UBSan reports this as undefined behavior. I think that the reason is as follows: C99 6.5.3.2#1 (Constraints) says: The operand of the unary & operator shall be either a function designator, the result of a [] or unary * operator, or an lvalue that designates an object that is not a bit-field and is not declared with the register storage-class specifier. But in case of an expression like "&array[-1]", the operand ("array[-1]") does not designate a valid object. Moral: check first -- assure that an expression is valid for a particular operation before carrying it out. Keep in mind that otherwise, the compiler is absolutely free to optimize out the *check*.

- Use the proper tile sequence when multiple pskies are present in a map. However, there's still only *one* psky chosen at map load time. - Handle three horizfrac cases: * 0, psky always at same level wrt screen * 65536, psky horiz follows camera horiz (e.g. E4L9) * otherwise, fall back to default hard-coded parallax implemented as an angle fraction

- Rename sky_t members: yscale -> horizfrac, bits -> lognumtiles. - Add default sky (8 tiles, horizfrac=32768 (i.e. 1/2 the scene horiz), offsets all zero) and CLOUDYOCEAN sky (8 tiles, horizfrac=65536, offsets all zero) to multipsky[]. - Get rid of "psky_t g_psky", merely maintaining a g_pskyidx instead. Set it up at map load time so as to keep the behavior of the legacy per-map psky: the last sector index with a matching psky ceiling wins. - In mapstate_t, save g_pskyidx too, not (former) pskybits and pskyoffs[]. - Make on-map-load global psky setup consistent for the game and editor by factoring it out into common.c: G_SetupGlobalPsky(). - Remove a couple of useless initializations, add some static assertions. This commit is more likely to introduce subtle differences in behavior. Specifically, getpsky() now always returns the default sky properties instead of the global sky ones (but with all-zero offsets) when no match for a suiting multi-psky is found. This is only likely to affect the yscale/horizfrac of non-multi-pskies when a global non-default multi-psky has been set up. Bump BYTEVERSION again.

- Consolidate psky* arrays into a "typedef struct psky_t" "g_psky" and "multipsky[]". - Factor out getting parallaxed sky properties into getpsky(). - Condense initial multi-psky setup by memcpy'ing from multipsky[]. - New function: MultiPsky_TileToIdx(). - Add new define PSKYOFF_MAX and related consistency-checking assertions. - Lower MAXPSKYTILES to 8 to reflect current usage (was 256). - Game: make multi-pskies consider dynamically-remapped MOONSKY1, BIGORBIT1 and LA. (Not very useful as the editor will still only act for the static values -- 80, 84 and 89, respectively.) An attempt has been made to preserve behavior even in strange cases, so this commit is unlikely to introduce regressions. Because of point 6, BYTEVERSION had to be bumped.

The old extension is an artifact from when I was thought that extensions should label whether a Lunatic module is to be used in internal/external contexts. Also, change test.lua slightly and commit a little updated documentation.

Its members are already laid out on the natural alignment boundaries. Compilation tested with {32-bit, 64-bit} x {Lunatic, non-Lua} builds.

It didn't turn out that useful and was giving a warning with one of the preceding changes.

Also, very slightly tweak a factor toward a "brighter" (farther visible) scene overall to account for the "circular" nature of the fragment distance (as opposed to an "ortho" distance in Build).

It was only ever used as upper bound to the time that a visibility change decays, but since it does that in an exponential fashion, there's really no point.

Now, if p->visibility differs from the constant visibility, the former converges toward the latter by three-quartering the difference each second totalclock increment (1/60th second). Negative player visibilities are not handled consistently throughout the different renderers, so make it look the same as 0 (can view to infinity).

So there's no "jump" from vis!=0 to vis=0 making the texture appear brighter.

Also override LTO to 0 in Lunatic builds.

Also, in test.elua, add a currently failing must-fail test (i.e. it wrongly succeeds where it should not!).

Because that function might have inserted tsprites itself, e.g. for the fake floor shadow. Also add one "spritesortcnt < MAXSPRITESONSCREEN" bound check in legacy ROR code.

That is, for shade of objects, use shade tables. For visibility, use GL fog. Mixing these two does not produce satisfactory results in areas where both shade and visibility darkening are high, such as around the E1L1 restroom.

- 'copy' may have them as source, 'getarraysize' works fine - 'setarray' and 'resizearray' are obviously forbidden - 'readarrayfromfile' and 'writearrayfromfile' is not implemented for them (use case?) Also, in the Makefile: don't use realpath on directories. It may give an empty string.

Add doc/how_to_build_lunatic.txt and doc/lpeg-lunatic.patch.

Don't yet make this official API though, since there are unresolved issues with newly created tsprites potentially being fed back to the animation loop. Add xmath.angvec(), xmath.bangvec(), tspr:set_sectnum(), tspr:setpos().

Having to tag these few variables and functions as ATTRIBUTE((used)) seems sensible, but what follows is just strange. OSD_Printf? headspritesect?

- swap 2d and 3rd args in xmath.rotate() -- now rotate(point, ang, pivot) - add vec3 method 'rotate', calling xmath.rotate - store game tic count in savegames

- some renames of functions/methods - con.actorvar: explicitly allow (for now) boolean and number values - added gv.gametic, player.holdskey(), player.all(), actor.check*()/hit*(), gv.rendmode, gv.REND, <bitint>:mask(), spr:getheightofs() - make read of nil var in "finalized" (live) module produce error, too - add test/helixspawner.lua

E.g. con.action{0, 4, delay=20}; for convenience, viewtype and incval are initialized to 1 (not 0) here.

That is, its value was only referenced during the duration of a function call that had previously set it. It was also never accessible from CON.

Using macros like AC_COUNT(t). Clean up related code: - comment out the dead condition noted in the previous commit - Lunatic: use get_count() instead of get_t_data(0) in one place, rename to _get_t_data(), i.e. make that method internal

... due to making hi/lotag unsigned in r3174. This commit keeps the compiler warning, the next one will remove it.