[Erry91]

What this changes

M1: photometric files become usable.

• luxpath-photometry reads IES LM-63 (1986, 1991, 1995, 2002, 2019; type C; relative and absolute photometry; TILT=INCLUDE) and EULUMDAT (all five symmetry indicators, Latin-1, decimal commas). Both are read into one full C-γ table in cd/klm.
• Symmetry expansion works for rotational, C0–C180, C90–C270, quadrant and full distributions. Outside the measured γ range the table gives 0 (no invented light).
• Emergency flux is set from a ballast lumen factor in (0, 1] or from lumens. Values out of range are errors.
• Errors name their line. Unsupported variants are refused with that line, never guessed: type A/B, TILT=<file>, several lamp sets.
• luxpath photometry <file> shows the format, the flux, the symmetry, the peak intensity and the light output ratio.
• cargo xtask fuzz [seconds] runs the mutation fuzzer.

Existing crates were evaluated first (docs/architecture.md):

• eulumdat and gldf-rs are AGPL and l3d_rs is GPL, so they are excluded.
• lidrs (MIT) is dormant since 2022. It panics on mutated input, has todo!() for type A/B files, and accepts truncated files as valid.

How you verified it

• cargo xtask ci, natively (macOS) and in the local Linux container. Merged on the local checks (hosted CI unavailable).
• Exit criteria:
  • synthetic files of every symmetry read identically from IES and LDT (tests/formats.rs, plus property tests);
  • cargo xtask fuzz 600 tried 22,847,141 inputs without a panic.
• Watched the tests fail:
  • a broken C90–C270 mirror turns formats.rs red;
  • a planted panic is found by fuzz_briefly.
• Real files: the two public sample files bundled with lidrs both read. The LDT's integrated light output (100.1 %) matches its declared value (100 %).
• The light output in the end-to-end test is checked by hand: the hand-written downlight gives 84.5 % from the zonal sums.

Life safety (AGENTS.md §2)

• No path reports "pass" when data is missing or ambiguous. Missing emergency output has no default, and unsupported variants are refused.
• Assumptions are documented: zero light outside the measured γ range, and the edition-specific factors are unverified against the texts (gap 14).

Never panic (AGENTS.md §3)

• No new unwrap/expect/panic!/indexing on input outside tests; NaN/inf guarded; allocations capped (16 MB files, MAX_TABLE_VALUES, 10,000 tilt angles)
• Fuzzed: 2-second run on every test, 10-minute run done

Provenance (AGENTS.md §1)

• No text, tables or figures from a standard
• Nothing from closed tools; no GPL/AGPL code
• Only synthetic fixtures committed (listed in crates/photometry/tests/fixtures/README.md)

Records

• parity/features.toml: symmetry, emergency flux shipped; IES and LDT readers partial (no real manufacturer files yet)
• docs/status.md, docs/gaps.md (rows 1–2 closed, 12–15 added), docs/roadmap.md, ROADMAP.md log