Fix Land's Edge — confirm existing CostPaidObjectMatchesFilter building block already handles it (test-only)

[rykerwilliams]

Summary

Closes a stale misparse-backlog entry for Land's Edge (World Enchantment, Chronicles 1995) with zero production code changes — this is a test-only PR.

Real Oracle text (Scryfall, confirmed this session): "Discard a card: If the discarded card was a land card, this enchantment deals 2 damage to target player or planeswalker. Any player may activate this ability."

The card was listed under docs/parser-misparse-backlog.md's root cause #2 ("Dropped intervening-if / gating condition"). Investigation found this is a stale backlog entry, not a live bug: the existing AbilityCondition::CostPaidObjectMatchesFilter building block (added 2026-05-04, previously only exercised in ConditionInstead-wrapped form by Agency Coroner / Surtland Flinger / Stormscale Anarch / Grab the Prize) already correctly handles Land's Edge's bare (non-instead) composition end-to-end:

• Condition extraction: "the discarded card was a land card" → CostPaidObjectMatchesFilter{filter: Typed([Card, Land])}
• Chunk-boundary protection for the leading "If ..., " clause
• Runtime evaluation via the discard cost's LKI snapshot (CR 608.2k / 400.7j)
• A separate, independently-existing PlayerFilter::All mechanism for "Any player may activate this ability" (CR 602.2)

This PR adds tests proving the claim, closing the coverage gap — it does not change any production behavior.

Files changed

• crates/engine/src/parser/oracle_effect/tests.rs — 2 new parser unit tests locking the full parse shape
• crates/engine/tests/integration/lands_edge_discard_land_condition.rs (new) — 5 GameRunner integration tests
• crates/engine/tests/integration/main.rs — mod line
• docs/parser-misparse-backlog.md — removed Land's Edge from root cause chore: update coverage stats and badges #2 (606→605 cards; totals rebased to 4760 distinct / 4794 total)

Gate A

$ ./scripts/check-parser-combinators.sh
(exit 0, no output — no violations)

Anchored on

• crates/engine/src/parser/oracle_effect/conditions.rs (parse_cost_paid_object_matches_filter_condition / parse_cost_paid_object_definite_noun_form) — the pre-existing parser this PR's tests exercise, unmodified.
• crates/engine/tests/integration/agency_coroner_suspected_cost_paid.rs — the existing sibling integration test the new lands_edge_discard_land_condition.rs is modeled on (same GameScenario/GameRunner pattern for a cost-paid-object-conditioned activated ability).

CR references

• CR 602.1 / 602.1a — activated ability cost paid by the activating player
• CR 602.2 / 602.2a — activation permission ("unless the object specifically says otherwise")
• CR 118.1 — costs
• CR 608.2c / 608.2k — resolution-time conditional; effect referring to a cost-referenced object
• CR 400.7j — cost-caused zone change makes the object findable by the effect
• CR 117.3d — priority (relevant to the non-controller-activation test fixtures)

Track

Developer

LLM

Model: claude-sonnet-5
Thinking: high

Verification

• cargo fmt --all — clean
• cargo clippy -p engine --all-targets -- -D warnings — clean, zero warnings (confirmed post-rebase)
• ./scripts/check-parser-combinators.sh — clean, exit 0 (confirmed post-rebase)
• cargo test -p engine --lib -- lands_edge — 2 passed, 0 failed (confirmed post-rebase)
• cargo test -p engine --test integration -- lands_edge — 5 passed, 0 failed (confirmed post-rebase)
• Discriminating-test coverage: land discard → 2 damage; nonland discard → 0 damage, ability still resolves (not short-circuited); the discard snapshot binds the specifically chosen object even when both a land and nonland are present in hand simultaneously; a non-controller can activate and pays the cost from their own hand (CR 602.1a); a sibling card without "any player may activate" correctly rejects non-controller activation, with the activating player genuinely holding priority in both fixtures so the rejection is a real activator_filter gate, not a vacuous no-priority artifact.
• All 9 CR citations verified against docs/MagicCompRules.txt directly.

Validation NOT performed (disclosed honestly)

An independent /review-impl pass was launched against this diff but ran for well over an hour without completing or responding to a mid-run status check, far exceeding every other review pass in this session's pipeline (which typically completed in 5–15 minutes). Rather than block indefinitely, I proceeded to ship without that independent review round. In its place: the implementing agent's own thorough self-verification (both new test suites passing, cargo fmt/clippy/check-parser-combinators.sh all clean, all 9 CR citations individually verified against the Comprehensive Rules text), plus this PR author's own post-rebase re-confirmation of every one of those checks from a clean rebase onto current main. Given the change is test-only (zero production code touched) and both new test suites independently exercise the real parse + real apply()/activation pipeline (not shortcuts), the residual risk from the skipped independent review is judged low, but I am not claiming a review happened that did not finish.

Scope Expansion

None.

Validation Failures

None (the skipped /review-impl pass is disclosed above under Validation NOT performed, not as a failure of a check that ran and failed).

CI Failures

None.

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[matthewevans]

Verdict: does not clear the test-only value bar — recommend closing (or reducing to a single assertion on the existing sibling test). This is a 573-line standalone test-only PR for a card that already works, exercising a building block that already has coverage.

🔴 Blocker — no regression guarded; the coverage already exists

By your own (accurate) investigation this is a stale backlog entry, not a live bug — Land's Edge already parses and resolves correctly with zero production changes. The bar for a standalone test-only PR is high because the suite is already very large: a new test earns its place only when it covers a real, otherwise-uncovered risk.

The AbilityCondition::CostPaidObjectMatchesFilter building block this PR exercises is already covered by four sibling integration tests — Agency Coroner, Surtland Flinger, Stormscale Anarch, Grab the Prize (crates/engine/tests/integration/agency_coroner_suspected_cost_paid.rs and siblings). Adding a fifth card's 427-line integration file (lands_edge_discard_land_condition.rs) plus 146 parser-test lines for the same already-covered condition is duplicative coverage — real cost to suite runtime and maintenance, near-zero marginal risk reduction.

🟡 The one kernel worth keeping (if it's real)

Your writeup notes the building block was previously exercised only in ConditionInstead-wrapped form, and Land's Edge is the bare (non-instead) composition. If that bare path is a genuinely distinct, currently-untested branch, it is worth one assertion — folded into the existing agency_coroner_suspected_cost_paid.rs sibling (or a single parser unit test locking the bare parse shape), not a new 427-line GameRunner file.

✅ Credit

The investigation is thorough and correct: Oracle text verified, CR annotations (602.1a / 608.2k / 400.7j) accurate, mod line registered, and correctly diagnosed as a non-bug. The diligence isn't in question — the coverage-value is.

Recommendation: close this unless the bare (non-instead) CostPaidObjectMatchesFilter path is genuinely uncovered; if it is, reduce to a single discriminating assertion on the existing sibling test rather than a new integration file.

[github-actions]

Parse changes introduced by this PR · 12 card(s), 10 signature(s) (baseline: main b8b23c9b9c00)

4 card(s) · static/Continuous · added: Continuous (affects=another you control permanent, mods=grant Hexproof)

Examples: Hollowhenge Huntmaster, Lecture Hall, Privileged Position (+1 more)

4 card(s) · static/Continuous · removed: Continuous (affects=another you control permanent, mods=grant Hexproof)

Examples: Hollowhenge Huntmaster, Lecture Hall, Privileged Position (+1 more)

3 card(s) · static/Continuous · field affects: you control creature Permanent → you control permanent

Examples: Cursed Wombat, Mishra, Tamer of Mak Fawa, Wondrous Crucible

2 card(s) · static/Continuous · added: Continuous (affects=another you control permanent, mods=grant Indestructible)

Examples: Avacyn, Angel of Hope, The Walls of Ba Sing Se

2 card(s) · static/Continuous · removed: Continuous (affects=another you control permanent, mods=grant Indestructible)

Examples: Avacyn, Angel of Hope, The Walls of Ba Sing Se

1 card(s) · static/Continuous · added: Continuous (affects=another you control permanent)

Examples: A Realm Reborn

1 card(s) · static/Continuous · added: Continuous (affects=transformed another you control permanent, mods=grant Trample, grant Ward)

Examples: Gargantuan Slabhorn

1 card(s) · static/Continuous · field affects: 1+ any counters you control creature Permanent → 1+ any counters you control permanent

Examples: Innkeeper's Talent

1 card(s) · static/Continuous · removed: Continuous (affects=another you control permanent)

Examples: A Realm Reborn

1 card(s) · static/Continuous · removed: Continuous (affects=transformed another you control permanent, mods=grant Trample, grant Ward)

Examples: Gargantuan Slabhorn

1 card(s) had Oracle-text changes (errata/reprint) — excluded as non-parser.

[rykerwilliams]

Thanks for the review — agreed, and pushed a fix.

Reduced from the 573-line standalone submission down to a single retained parser test (lands_edge_discard_land_condition_parses_as_bare_intervening_if), which locks the full bare parse shape in one assertion block:

• Removed crates/engine/tests/integration/lands_edge_discard_land_condition.rs in full (427 lines, 5 tests) and its main.rs mod line — the runtime path (CostPaidObjectMatchesFilter evaluation) is already exercised by the four existing sibling integration tests you cited.
• Removed the second parser test (lands_edge_without_any_player_clause_has_no_activator_filter) — it was proving PlayerFilter::All, a separately-established, already-generically-tested mechanism, not the bare-condition distinction.
• Kept the one test that proves the genuinely new thing: Land's Edge's condition is a bare (non-ConditionInstead) composition, unlike all four existing siblings. At runtime this isn't actually a new code path — evaluate_condition's CostPaidObjectMatchesFilter arm fires identically whether or not it's wrapped in ConditionInstead — so the distinction is provable at the parser level alone, matching your suggested "or a single parser unit test locking the bare parse shape" option.

One disclosure: I could not get a fresh local cargo test run to complete after this reduction — 5 consecutive attempts were killed mid-compile (this fork's shared cargo build lock was held by another concurrent agent throughout). This commit is a pure deletion with zero logic changes; the one retained test is byte-identical to its previously-verified-passing form from the original submission. CI will confirm.

[matthewevans]

✅ Approve — the slimmed revision clears the value bar; it now locks a genuinely uncovered parse path.

My prior review declined the +573 version as duplicative (value-bar): CostPaidObjectMatchesFilter was already covered by four sibling tests. The +117/−4 revision answers that directly by targeting the one composition those siblings don't exercise.

✅ Clean

• Bare vs. wrapped is a real, distinct parse path. Every existing CostPaidObjectMatchesFilter test in oracle_effect/tests.rs asserts it wrapped in ConditionInstead (Stormscale Anarch, Surtland Flinger, Witch's Oven). Land's Edge is the bare intervening-if form (CR 608.2c) — the condition gates the whole effect, not a "…instead" override. Grep-confirmed: no sibling exercises the unwrapped AbilityCondition::CostPaidObjectMatchesFilter.
• Discriminating and non-vacuous. The reach-guard (abilities.len() == 1) blocks a degenerate pass; the let AbilityCondition::CostPaidObjectMatchesFilter { .. } = cond else { panic! } would fail on the exact condition: null regression this card sat under in parser-misparse-backlog.md — which the PR correctly removes from that list.
• Locks the full shape, not just the condition: unfiltered Discard cost, DealDamage(2) to a player-or-planeswalker Or, and activator_filter == Some(PlayerFilter::All) for "Any player may activate this ability."

Recommendation: approve and enqueue. Test-only, CLEAN, fully green (both Rust shards + coverage-gate + parser gate). The coverage is unique and load-bearing.