What changed in EDuke32, build by build

- Return early from drawsprite_classic() if tspr->owner if oob. Commonize that check between renderers into bad_tspr(). - Make the BIT() macro a left-shift of the *unsigned* number 1, preventing expansion to (1<<31).

In those Clang versions, -fcatch-undefined-behavior was replaced by -fsanitize=<what>. For Clang 3.2, we enable the address sanitizer and for 3.3 also the miscellaneous undefined behavior one. Currently, there's no version auto-detection from the Makefiles, it's set to 3.3 and the others are commented out.

Numbers 0 and 1 mean 'no spread', but the latter does one krand() call. Negative numbers are reserved for potential future use.

... four lines later. I don't know if that is the intended semantics for this function, but nasal demons is clearly not what we want. Please review!

Also, make G_LoadAddon() local to game.c.

LameDuke's shade table has 32 gradients of shade, like Duke3D. For the translucency table though, only the diagonal + one half is stored because it's symmetric (50/50 translucency).

grpinfo { name "YourGameName" scriptname "yourgame.con" defname "yourgame.def" size yourgrpsize dependency DUKE15_CRC flags GAMEFLAG_ADDON crc yourgrpcrc } The "flags" field can also take the value "GAMEFLAG_DUKE" to define a new base game. Since this is the first commit of a new feature, certain semantics of the implementation and syntax are liable to change as necessary.

Set it to >=0 to override the base palette (water, slime, ...) in 3D mode. This can be useful to navigate Lameduke's underwater sections, since its LOOKUP.DAT only contains 8 lookup tables, but no other *base* palettes. Consequently, base palettes >=1 will make the whole screen black when Lameduke's LOOKUP.DAT is loaded into EDuke32 or Mapster32.

Also remove the two instances of commented out code, but not the respective CON structure access code.

This fixes the bug reported here: http://forums.duke4.net/topic/5340-bug-with-masked-walls/page__view__findpost__p__123233 But I haven't checked the change for any adverse effects.

This fixes RESPAWN preview wrongly clamping the previewed tsprite z coords to [-524288 .. 524288] (x/y limits). Also, make the previewed tsprites be 33% translucent with the option of 66% translucency by setting 'showrespawn_fulltrans' to 1 in a.m32.

- colorize the trailing part if there is corruption or (new) if sprites were removed - also corrupt-check when loading <V7 map - move start{pos,ang,sectnum} to build.c - remove pointless ...[].extra = -1 before loading map, scantoasc*[] comment

- In the A_MoveSprite() code that transports projectiles due to an SE7 (introduced in r1450 / legacy ROR), only report "success" if the transportation succeeded. - Clear newly introduced internal SPRITE_DIDNOSE7WATER flag after checking it.

Also, in two cases, don't assign A_IncurDamage() result to a variable if it's not needed.

- In S_PlaySound(), move the sound index bound check above an indexing. - For A_CheckHitSprite(), and A_FindPlayer(), allow NULL second arg. - In A_ShootWithZvel(), make some one-letter vars be int32_t, making storing safeldist() results in them meaningful. - In MaybeTrainKillEnemies(), remove two redundant checks and move another one further up. - Comment that SIDEBOLT1 will never be translucent as was probably intended. - In G_MoveFX(), fix an always-true comparison.

[1] http://forums.duke4.net/topic/2961-paper-cuts-minor-bugs-and-annoyances/page__view__findpost__p__153060

Let me know if this breaks any mods that have #2462 with an x-dimension other than 320.

(Note: We are still vulnerable to crashes here (and likely in other places), but you have (4096-16) extra characters to work with. We should consider replacing many uses of Bsprintf with Bsnprintf.)

This fixes liztroops not passing TROR layers when on the jetpack, as well as other enemies capable of moving upward, and also COMMANDER and DRONE (for which the code is special-cased).

Also, sync MAXBUNCHES to new limit in lunatic/defs_common.lua.

... like was done with clipmove() previously. Also, do some code clean up. There are no changes of functionality.

Nag: The behavoir of atoi() is *undefined* if "the value of the result cannot be represented" (C99 7.20.1). Note: my nagging about undefined behaviors will *not* get less over time. :P

The Lunatic build compiles with new structures for sector and wall types. They have separate members for TROR {up,down}{bunch,nextwall}, so there are no conflicts with other uses of members into which they were previously shoehorned. Also, the maximum bunch limit is bumped to 512 in that build. Currently, loading from V7/8/9 and saving to V7 and V8 are supported.