test(dwarf): LS-CP-4 aliases + gate scans integration tests

CHANGELOG.md

@@ -6,6 +6,24 @@ All notable changes to this project will be documented in this file.
 
 ### Fixed
 
+- **LS-CP-4 regression coverage + LS-N gate integration-test scan**
+  (`tools/run_ls_verification.py`,
+  `meld-core/tests/dwarf_strip.rs`). The LS-N verification gate was
+  previously invoking `cargo test --lib`, which filtered out
+  integration tests under `<package>/tests/`. Three pre-existing
+  tests in `dwarf_strip.rs` (`default_strips_dwarf`,
+  `passthrough_preserves_dwarf`, `default_is_strip`) already pin
+  LS-CP-4 ("DWARF passthrough emits address-incorrect debug info on
+  fused output") via Phase 1.5's Strip-default policy, but the gate
+  reported the entry as missing because the test names didn't match
+  the `ls_cp_4_*` convention AND the integration-test binary
+  wasn't being scanned. Adds three convention aliases delegating to
+  the existing test bodies, and drops `--lib` from the gate runner
+  so both lib and integration tests participate. Gate verdict moves
+  from 16/19 verified to **17/19 verified**; remaining
+  missing-bucket entries are LS-A-8 and LS-A-9 (both need
+  net-new tests, not aliases).
+
 - **LS-A-19 regression coverage** (`meld-core/src/merger.rs`). PR #156
   fixed the `imp.name.ends_with(rn)` suffix-collision bug
   (`float` / `bigfloat` cross-resource confusion) but landed without

meld-core/tests/dwarf_strip.rs

@@ -115,3 +115,30 @@ fn default_is_strip() {
     let cfg = FuserConfig::default();
     assert_eq!(cfg.dwarf_handling, DwarfHandling::Strip);
 }
+
+// LS-N verification gate convention aliases. These delegate to the
+// existing tests above; their purpose is the discoverable
+// `ls_cp_4_*` name that `tools/run_ls_verification.py` matches
+// against. The original tests stay in place as the canonical bodies
+// (preserves git blame + grep continuity).
+//
+// LS-CP-4 ("DWARF passthrough emits address-incorrect debug info on
+// fused output") is the loss scenario the Strip-default policy
+// directly mitigates — Phase 1.5 made Strip the default so the
+// broken passthrough is no longer the silent fallback. These tests
+// pin that policy.
+
+#[test]
+fn ls_cp_4_default_strips_dwarf() {
+    default_strips_dwarf();
+}
+
+#[test]
+fn ls_cp_4_passthrough_preserves_dwarf_explicitly() {
+    passthrough_preserves_dwarf();
+}
+
+#[test]
+fn ls_cp_4_default_is_strip() {
+    default_is_strip();
+}

tools/run_ls_verification.py

@@ -76,25 +76,33 @@ def load_approved_ids(yaml_path: Path) -> list[str]:
 
 
 def run_prefix(package: str, prefix: str) -> tuple[int, int, int]:
-    """Return (matched, passed, failed) for `cargo test --lib <prefix>`.
+    """Return (matched, passed, failed) for `cargo test <prefix>`.
 
     `matched` = total selected tests (passed + failed + ignored). When
     cargo finds no matching tests, it reports `0 passed; 0 failed; 0
     ignored` and still exits 0; the caller treats matched=0 as missing.
+
+    Drops the `--lib` filter so integration tests under
+    `<package>/tests/` are also scanned. Several LS-N scenarios pin
+    their regression coverage in integration tests (e.g. LS-CP-4 in
+    `dwarf_strip.rs`); excluding them caused the gate to misreport
+    those entries as missing.
     """
     cmd = [
         "cargo",
         "test",
         "-p",
         package,
-        "--lib",
         "--no-fail-fast",
         "--",
-        "--exact" if False else "--test-threads=1",
+        "--test-threads=1",
         prefix,
     ]
     proc = subprocess.run(cmd, capture_output=True, text=True)
     passed = failed = ignored = 0
+    # cargo emits one `test result:` line per target (lib + each
+    # integration test binary), so we sum across all of them. A
+    # prefix that matches in *any* target counts as "matched".
     for line in proc.stdout.splitlines():
         m = _TEST_LINE.match(line)
         if m: