[Testing] Validate SDL-event → SFML-enum input parity (asserted "Full", never tested) #389

New issue

Open

opened 2026-07-16 16:47:42 +00:00 by john · 0 comments

john commented

2026-07-16 16:47:42 +00:00

Owner

Scope

Enumerate the SDL-event → SFML-enum translation in the SDL2 backend (src/platform/SDL2Types.h / event pump) and cross-check it against the desktop sf::Keyboard/sf::Mouse enums the game code consumes.
Add a regression test asserting full coverage: every mcrfpy.Key / mcrfpy.MouseButton value the Python API exposes has a defined SDL source mapping (and vice versa — no SDL event that lands as an invalid/zero enum).
Cover modifiers (shift/ctrl/alt), mouse buttons, and any key the demos/snippets actually use.
Where a mapping is genuinely absent (no SDL equivalent), document it in the parity matrix rather than leaving it implicit.

Notes

The enum-translation table is a pure data structure — much of this is testable headlessly against the mapping itself, without a live browser.
Any gap found becomes a granular follow-up (system:input).

Part of #387. CLAUDE.md's Web Build Constraints table marks Input "✅ Full — SDL events translated to SFML enums", but nothing validates that translation table. An unmapped key or modifier would silently do nothing on web while working on desktop. ## Scope 1. Enumerate the SDL-event → SFML-enum translation in the SDL2 backend (`src/platform/SDL2Types.h` / event pump) and cross-check it against the desktop `sf::Keyboard`/`sf::Mouse` enums the game code consumes. 2. Add a regression test asserting full coverage: every `mcrfpy.Key` / `mcrfpy.MouseButton` value the Python API exposes has a defined SDL source mapping (and vice versa — no SDL event that lands as an invalid/zero enum). 3. Cover modifiers (shift/ctrl/alt), mouse buttons, and any key the demos/snippets actually use. 4. Where a mapping is genuinely absent (no SDL equivalent), document it in the parity matrix rather than leaving it implicit. ## Notes - The enum-translation table is a pure data structure — much of this is testable headlessly against the mapping itself, without a live browser. - Any gap found becomes a granular follow-up (`system:input`).