[pogoflow]

PpuDrawWholeLine implements modes 0-3 and treats every other mode as mode 7. With NewRenderer on, a mode 4, 5 or 6 screen therefore samples BG1's tilemap as a mode-7 pixel map. I hit this on a HiROM title: a mode 5 menu (hi-res text over a 4bpp background) came out as vertical colour stripes, while VRAM, CGRAM and OAM matched a reference emulator. NewRenderer = 1 is the host default, and widescreen forces it, so the reference renderer is not a way out there.

Change
ppu_runLine routes modes 4-6 to ppu_draw_whole_line_legacy, which already handles 16-pixel-wide tiles, the hi-res column split and offset-per-tile. Modes 0-3 and 7 stay on the optimized path. This is meant as a correct fallback until the optimized renderer grows those modes, not as a replacement for that work.

Test
tests/ppu/ppu_hires_mode_test.c (no ROM needed): for modes 4, 5 and 6 the optimized renderer must produce the compatibility renderer's scanline, mode 5 must show the 16-pixel tile split, and mode 1 must stay on the optimized path. It is registered in tests/ppu/run.ps1 and tests/run_c_tests.sh.

Against unchanged ppu.c, 6 checks fail (modes 4/5/6 differ from the compatibility scanline, and the mode 5 tile split is missing). With the change it passes. tests/ppu/run.ps1 passes all six PPU tests, and the compiler warning count for ppu.c is unchanged. Full Python suite on the branch: 846 passed, 2 skipped (Windows, MinGW gcc 16.2).

On Windows, tests/run_c_tests.sh passes every section up to and including the PPU ones, then stops at "SNES lobby caps" because setenv/unsetenv don't exist on MinGW. That test and its section are untouched by this change.

[mstan]

No blocking code findings. The per-scanline mode check routes only the unsupported modes 4/5/6 to the existing compatibility renderer; modes 0–3 and 7 retain the optimized path. The fallback preserves the compatibility renderer's existing 256-sample hi-res presentation; it does not claim full 512-pixel output.

Reviewed with #199 and #214 on current main 246c0d4. All six production PPU tests pass, including the new regression. Extra compiled checks compare the fallback at 256/342/512/800-pixel output widths and verify framebuffer guards. The combined tree also passes the codegen width lint, 860 Python tests (2 platform skips), and the production runtime dispatch fixture. All three workflows on this reviewed PR head are green.

The shared regression branch is review/open-prs-final-20261026 (b19edfb). This is code and automated-fixture approval; the new #214/#215 pair still needs the normal combined title gameplay/current-build save-load/runtime gate and owner handoff before merge. The fixtures alone do not establish real-title menu appearance or frame times. No games were relaunched in this review.

[mstan]

Combined regression follow-up: this reviewed head is included with #199/#214 on current-main integration branch review/open-prs-final-20261026 at b19edfb. All 860 Python tests (2 platform skips), six production PPU harnesses, runtime dispatch and extra 256/342/512/800-pixel framebuffer-bound checks pass. Existing 256-sample hi-res presentation remains the supported limitation.

All eight matched baseline/candidate Release builds are provenance-stamped. Muted gameplay, movement, current-build save/load and nonzero emulated audio checks pass in Mega Man X, ALttP, SMW and Super Metroid. Brief built-runtime checks show no consistent introduced regression, including opposite-order SM repeats after an initial timing outlier. SMW's bounded Yoshi's House check remains paced near 60fps and does not establish wide-stage margins; SM's original Zebes-loader setup bypasses Ceres.

The automated combined gate is complete. Approval stands; the owner's quick visual/play check of the new #214/#215 batch remains pending before merge. No author changes requested.