fix(dwarf): remap DW_AT_high_pc lengths to the fused layout (#319 inc 1)

[avrabe]

What

Fixes the dominant cause of #319 — meld fuse producing DWARF that fails llvm-dwarfdump --verify with hundreds of out-of-parent + overlapping DIE ranges.

Root cause

gimli routes only address-form values through convert_address; DW_AT_high_pc encoded as a length (DW_FORM_udata, the common Rust/LLVM form) is a constant it copies verbatim. Fused bodies change length (LEB re-encode of the merged index space), so every subprogram/lexical-block/inlined DIE became [remapped_low, remapped_low + STALE_input_len) — overrunning its parent (grew) or the next function (shrank).

Fix

After the gimli conversion, walk the write DIEs and rewrite each low_pc+high_pc-length pair via a new AddressRemap::corrected_high_pc: invert the remap for low_pc within its output body, then use the span's output_body_end when the range ends at the function end (avoids the aliased-boundary translate() None) or translate for interior ends. Tombstoned (dropped) low_pc is left untouched — llvm-dwarfdump ignores those DIEs.

Measured (records+variants repro)

	before	after
overlapping	480	0
not-contained	534	11
invalid-expr / invalid-loc	18 / 17	18 / 17
total	1049	46

Scope

Inc 1 clears ~96% (all overlapping + 98% of out-of-parent). Residual is distinct mechanisms tracked as follow-ups on #319: nested DW_AT_ranges on inlined/lexical DIEs (11/4 not-contained), and the .debug_loc tombstone-vs-base-selection collision (Cause #2, invalid-expr/loc).

Test: corrected_high_pc_rewrites_length_to_fused_layout (grew/shrank/interior/tombstone) — deterministic, no llvm-dwarfdump dependency. 24 dwarf tests pass.

Refs: #319.

🤖 Generated with Claude Code

[github-actions]

LS-N verification gate

✅ 58/58 approved LS entries verified

	count
Passed (≥1 test, all green)	58
Failed (≥1 test failure)	0
Missing (no ls_*_NN_* test found)	0

_{Approved loss-scenarios.yaml entries are expected to have a
regression test named ls_<letter>_<num>_* (e.g. LS-A-11 →
ls_a_11_*). The gate runs each prefix via cargo test --lib --no-fail-fast and aggregates pass/fail/missing.}

Failed LS entries

(none)

Missing regression tests

(none)

_{Updated automatically by tools/post_verification_comment.py.
Source of truth: safety/stpa/loss-scenarios.yaml.}

[avrabe]

Mythos delta-pass — NO FINDINGS (meld-core/src/dwarf.rs)

Ran the discover protocol on this PR's dwarf.rs delta (a fresh adversarial session). Five hypotheses tested, four refuted empirically:

1. Wrong-span reverse map at a boundary — refuted: output bodies are non-overlapping, so a concrete out_low at a shared output_body_end == next.output_body_start boundary resolves to the next function's start (its own low_pc), not the prior's exclusive end. The input-space aliasing translate must tombstone doesn't exist in output space.
2. reverse_in_span mis-inversion — refuted: InstrOffset.new is strictly increasing (each operator ≥1 output byte), so the exact new == instr_new match is unambiguous; the region-1/2 split is the exact inverse of translate.
3. orig_end past the function — the one theoretical wrong-length path (a high_pc exceeding its body would translate into the next function). Unreachable from valid toolchains (subprograms end at input_end; nested blocks/inlines stay interior; CUs use DW_AT_ranges, not high_pc). Hardened anyway (commit 594ba57): bail when orig_end > span.input_end, so it's now provably impossible even on malformed input (LS-D-1).
4. Silent no-op on real DWARF (the key risk: clang/rustc emit high_pc as DW_FORM_data4, my match is on Udata) — refuted end-to-end: gimli normalizes the constant-class high_pc to Udata in the read→WriteDwarf::from path, so the match fires and the length is corrected. Verified with hand-assembled data4 bytes.
5. Mutation walk / adapter unit — refuted: correct_high_pc_lengths runs before append_adapter_unit, so the synthetic <meld-adapter> unit is never walked; the DFS collects ids then mutates per id (no cycles, each once).

No failing oracle for any wrong-attribution defect. Adding mythos-pass-done.

[github-actions]

Mythos delta-pass (auto)

✅ NO FINDINGS across 1 Tier-5 file(s)

File	Verdict	Hypothesis
``	✅ NO FINDINGS	—

_{Auto-run via anthropics/claude-code-action@v1
(SHA-pinned) on the touched Tier-5 files, using the
maintainer's Max-plan OAuth token. See
.github/workflows/mythos-auto.yml and
scripts/mythos/discover.md.}