[Cimos]

Fixes #426

IN-6 now covers the default output names of more EDA tools. The new classifier pass (other_eda_scheme in etchy-core/src/naming.rs) runs before the stem heuristics. It only matches extensions and stems that the tools document as defaults.

Scheme	Names added
Eagle legacy CAM	.cmp .sol .stc .sts .plc .pls .crc .crs
Eagle 9 templates	copper_top[_l1], copper_inner_l<n> (physical position → ordinal), copper_bottom[_l<n>], soldermask_*, solderpaste_*, silkscreen_*, profile.gbr
OrCAD Layout	.top .bot .smt .smb .sst .ssb, .ast → documentation
Allegro	TOP.art, BOTTOM.art, SOLDERMASK_*, SILKSCREEN_*, PASTEMASK_*; DRILL.art → documentation (previously read as the drill layer)
EasyEDA	Gerber_InnerLayer<n> (any ext, incl. plane .GP<n>), .GTA/.GBA → documentation
Protel	same extensions as Altium; pinned by its own corpus list, no new rules

These names stay Other on purpose, and the tests pin that: .drd (Eagle Excellon vs OrCAD drill drawing; content sniff still loads Excellon as drill), plane .gp<n> without the EasyEDA stem, Eagle .ly<n>, OrCAD .gnd/.pwr/.fab, Allegro user-named inner films, malformed copper_inner_l1.

Tests

• corpus/naming/<scheme>.txt lists <filename> => <kind> with source citations. tests/naming_corpus.rs checks every file.
• The altium.txt and kicad.txt baselines passed against the code before this change, so the existing classifications are unchanged.
• Per-scheme unit tests are in naming.rs.
• cargo fmt, clippy -D warnings, and cargo test --workspace --exclude etchy-gui all pass.

Docs: IN-6 is marked ✅, plus a CHANGELOG [Unreleased] entry and a note in corpus/README.md.

[Copilot]

🟡 Changes recommended

Layer detection still misclassifies files because the new scheme pass follows conflicting inner-copper heuristics and the DRILL substring check matches board-name prefixes, breaking rename-tolerant pairing.

2 open issues

• Run scheme layer detection before inner-copper heuristics · New
• Match DRILL as a film name, not a board-name substring · New

What changed in this PR

Extends etchy’s shared filename classifier with additional EDA output schemes, supporting layer identification and cross-scheme pairing for IN-6.

Changes:

• Adds Eagle, OrCAD Layout, Allegro and EasyEDA naming rules, including documentation classification for DRILL.art.
• Adds per-scheme corpus tests, with Altium/KiCad baselines and Protel compatibility cases.
• Updates requirements, corpus documentation and release notes.

File	Description
docs/​REQUIREMENTS.md	Marks IN-6 complete and documents supported schemes.
crates/​etchy-core/​tests/​naming_corpus.rs	Checks expected classifications across all corpus lists.
crates/​etchy-core/​src/​naming.rs	Adds scheme-specific classification rules and unit tests.
corpus/​README.md	Links to the filename-classification corpus.
corpus/​naming/​README.md	Explains corpus format and ambiguous-name handling.
corpus/​naming/​protel.txt	Pins Protel’s shared Altium extension mappings.
corpus/​naming/​orcad-allegro.txt	Covers OrCAD and Allegro default names.
corpus/​naming/​kicad.txt	Records existing KiCad classifications.
corpus/​naming/​easyeda.txt	Covers EasyEDA layer and drawing names.
corpus/​naming/​eagle.txt	Covers legacy CAM and Eagle 9 names.
corpus/​naming/​altium.txt	Records existing Altium classifications.
CHANGELOG.md	Announces expanded naming support and drill-film reclassification.

_{Review effort: Balanced}

[Copilot]

🔵 Needs a closer look

Layer-classification precedence and overly broad stem matching misclassify exports and accept conflicting copper-layer identifiers, breaking cross-scheme pairing and mask-layer gating.

2 open issues

• Match DRILL as a film name, not a board-name substring
• Run scheme layer detection before inner-copper heuristics

1 previously missed issue

In code that hasn't changed since last review

Validate EasyEDA stem before generic numeric extensions

crates/​etchy-core/​src/​naming.rs:135

The extension-agreement check is bypassed for .G<n>: classify("Gerber_InnerLayer2", "G3") returns InnerCopper(3) through the earlier generic extension branch, while the equivalent .GP3 mismatch returns Other. This can pair a conflicting EasyEDA export with the wrong copper layer. Validate recognized EasyEDA stems before the generic numeric-extension branch, explicitly return Other on disagreement, and add a mismatched .G3 regression case.

_{Review effort: Balanced}