SuperTuxKart: the big-endian port and the VRAM (2026-10-16 to 10-18)

Symptom. The research of 2026-10-16 had two warnings: the FAQ asks for at least 1 GB of RAM against our 217 MB, and Debian had an old powerpc colour bug (#714519). On 2026-10-17 Debian’s build stopped while loading: its mesh loader refuses big-endian (e708512). Our +ps3.264 loaded, the karts rendered, the track didn’t: VRAM exhausted (210 failed buffer allocations), the minimap FBO incomplete, a SIGSEGV at exit.

Cause and fix.

• 0001: the .spm mesh and armature loader byte-swaps the little-endian data instead of refusing to run (upstream #5578 was closed as not planned). All 1268 .spm files parse identically on amd64 and ppc64.
• 0002: textures and vertex colours in the right channel order on the OpenGL driver (the old powerpc colour bug).
• 0003: the legacy renderer no longer crashes when render-to-texture fails, and the minimap’s render target is power-of-two (512x512).
• 0004: texture compression reaches the legacy renderer: opaque power-of-two textures as DXT1, compressed by Mesa at upload.

With 0003-0004 (ps3.306) the VRAM went from 210 to 132 MB (98 MB with nohd or lean), HUD 13-19 fps. On 2026-10-18 the track drew (batch 2026-10-18-batch8): base 12.6 fps with 13.5 s of stalls, lean 21.4 with 2.4 s.