Cosa è cambiato in EDuke32, build per build

This allows the startup window to be used on Steam-based devices like the Steam Deck without switching to desktop mode.

It looks like there were a couple interpolation related fixes in libxmp since the last time we tried enabling this, so maybe it'll work without crackling and crashes this time.

Note: 'skill == 0' can be obtained from the command line by either * '-m': "monsters off", or * '-s0': 'ud.m_monsters_off' is *not* set

In 'GetCornerPoints()', fix condition for when to use 'get_floorspr_points()' to include sloped sprites as well.

(More precisely, in renderDrawMasks()'s check "does the maskwall separate the tsprite center from the camera?".) This provides more reasonable rendering order when the sprite (as drawn) is heavily biased to one side of the maskwall. Note [1]: mask/mask intersections are in general problematic since there is no further subdivision. For opaque objects, the OpenGL modes "work around" the issue thanks to the depth buffer. See: - 'mask_drawing_order.map', the STEAM sprites which by position are all on the other side of the right-hand side fence, but as drawn may or may not be more towards one particular side. (The topmost one with xoffset 50 is the first to be drawn behind it; note that 'picanm[STEAM].xofs == -47'.) - Issue #136, the first example screenshot. However, there, the fence intersects the STEAM sprite quite exactly in the middle (as drawn), so note [1] above applies.

- pull out repeated expression 'spr->ang + angofs' into local 'ang' - tweak 'editorDraw2dScreen()', too

Background: "search for named tile" does not work as expected.

Meaning the 'while (maskwallcnt) { ... loop over tpsrites ... }' block in 'renderDrawMasks()' that is not conditional on the rendering mode.

- 'PolymostPrepareMasks()' - 'PolymostDrawMasks()' - 'DrawDebugSpriteOutlines()' In passing: - Declare once and reuse 'isPolymost' local - In '#if 0'-d debug code, fix an inconsistency and a call by an obsolete function name

(NIFC: no intended functional changes)

source/mimalloc/src/options.c:228:3: runtime error: call to function mi_log(char*, void*) through pointer to incorrect function type 'void (*)(const char *, void *)' source/build/src/osd.cpp:1755:44: runtime error: null pointer passed as argument 2, which is declared to never be null source/build/src/glbuild.cpp:296:9: runtime error: load of value 4294967295, which is not a valid value for type 'glsamplertype' source/duke3d/src/gamestructures.cpp:355:125: runtime error: member access within null pointer of type 'DukePlayer_t' ---- Note: we are also getting e.g. > source/build/include/compat.h:1008:72: runtime error: store to misaligned address 0x(...) > for type 'uint32_t' (aka 'unsigned int'), which requires 4 byte alignment due to '-fsanitize=alignment' implied by '-fsanitize=undefined'. This is left as-is.

Exposed by UBSan, e.g.: source/build/src/engine.cpp:11090:67: runtime error: null pointer passed as argument 2, which is declared to never be null

- 'unsigned-integer-overflow': happens too frequently (and is not UB) - 'float-divide-by-zero': see https://reviews.llvm.org/D63793 and D64317

This is already covered separately by undefined parameter checks on label use and by keyword checks on label definition.