chore(release): v0.42.0 — trap-preservation VC + float completion + parity benchmark + #739 soundness hub

[avrabe]

Cuts v0.42.0 from the seven-lane hub (#736/#737/#738/#741/#742/#743/#744, all merged + main-train CI green).

Highlights

• VCR-VER-002 Phases A+B (Arc: verify what ships (close #73) — v0.12 control-flow model + division traps #166, ordeal#59): trap-preservation is a provable obligation — div/rem, memory OOB, call_indirect (3 traps, closed-world mode), unreachable, all six trunc variants (31-row thumb-2 f32: i32.trunc_f32_s/u silently saturate instead of trapping on NaN/Inf/out-of-range (bare VCVT, no guard) — WASM Core §4.3.3 soundness bug #709 boundary table). Red-first gate: every trap-drop caught (Sat+counterexample), every preserve accepted (Unsat, LRAT).
• --shadow-stack-size: static ABOVE sp_init is baked at its 1MB linmem offset, not re-based → OOB read in shrunk layout (post-link oracle wrongly passes) #739 soundness fix (fix(#739): static ABOVE sp_init — relocate (never bake) + de-vacuate the in-range oracle #744): baked above-sp_init statics relocated; the vacuously-green in-range oracle de-blinded; no-reloc shrink refuses.
• Falcon float completion (feat(vfp): f32-across-call spill/rehome (#719 residual) + scalar f64 phase 2 for falcon (#369) #741): f32-across-call + scalar f64 (m7dp), 187/187 + 126/126 vs wasmtime; caught 2 latent soundness bugs pre-shipping.
• Generated-not-mirrored Rocq model (feat(coq): VCR-ISA-001 #667 increment 2 — generate, don't mirror: VcrSelRules.v re-exports the generated model #738): rule_X := Gen.rule_X; 472 honest recount.
• Parity benchmark (feat(#735): reproducible wasm-AOT vs native-C parity benchmark — measured bytes, cited baselines, cycles OPEN #737, Parity benchmark: reproducible wasm→synth-AOT vs native-C→LLVM comparison — counter the 'wasm is ~40% slower' claim with measured cycles+bytes #735): gust_mix clamp 14 B vs 26 B measured gcc -Os = 0.54× of native, falsification command per row, cycles OPEN (gale).
• Mask elision (feat(vcr-perf): redundant-mask elision — proof-carrying #494 phase 3+ (VCR-PERF-002) #736): gust_kernel 34→20 B certified. Spill reduction (perf(#390): conditional-branch-transparent stack-reload forwarding — gust_poll 740→724 B (VCR-RA) #743): gust_poll 740→724 B, 26 shrink/0 grow.

Pin sweep 0.41.0→0.42.0 (incl. npm), claim gate 18/18. Cold-agent review runs against this PR before tag.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!