feat(dwarf): DwarfHandling::Remap end-to-end (#143 DWARF Phase 2 inc 3b)

CHANGELOG.md

@@ -4,6 +4,46 @@ All notable changes to this project will be documented in this file.
 
 ## [Unreleased]
 
+## [0.19.0] - 2026-05-29
+
+### Added
+
+- **`DwarfHandling::Remap` — end-to-end DWARF address remapping**
+  (#143 DWARF Phase 2 increment 3b, `meld-core/src/dwarf.rs`). The
+  final piece of DWARF Phase 2: reads an input core module's `.debug_*`
+  sections, translates every code address to the fused code section via
+  the `AddressRemap` engine (v0.18.0), and re-serializes a single
+  remapped DWARF set with `gimli::write::Dwarf::from`. New `gimli`
+  dependency. Exposed via `meld fuse --dwarf remap` (modes: `strip`
+  (default), `passthrough`, `remap`).
+  - The per-function instruction offset map is recovered **post-hoc**
+    by walking the input and final-output operator streams in lockstep
+    — no state threaded through the merge hot path, and it reflects
+    whatever rewriting actually happened (including the adapter-wiring
+    re-rewrite). A per-function operator-count or locals-prefix mismatch
+    aborts the remap (correct-or-strip).
+  - **Correct-or-strip throughout.** `gimli::write::Dwarf::from` is
+    all-or-nothing on addresses, which is used as the safety gate: only
+    the structurally-invariant code-section base (address 0) is
+    special-cased; any other unmapped address fails the conversion and
+    falls back to stripping rather than emitting a wrong address.
+  - **Single DWARF source** is supported in this increment. Inputs
+    where more than one core module carries DWARF fall back to `strip`
+    with a warning (merging independent DWARF unit sets is deferred);
+    zero DWARF sources is a no-op.
+  - Encoding uses a three-pass dance so the remapped `.debug_*` land in
+    the attestation/provenance-hashed bytes (trailing custom sections
+    don't shift code offsets, so the remap built from pass A is valid
+    for the final output).
+  - New loss scenario **LS-D-1** (wrong remapped DWARF address →
+    de-grounded downstream coverage/breakpoints) is `approved`, gated by
+    `dwarf::tests::ls_d_1_remap_translates_low_pc` (full gimli
+    read→convert→write→read oracle). **Residual:** `DW_AT_high_pc`
+    encoded as a *length* is copied verbatim, so a function's reported
+    byte length may be off by intra-function LEB drift; `low_pc` and the
+    line-number program (what debuggers and `pulseengine/witness` use)
+    are correct.
+
 ## [0.18.0] - 2026-05-29
 
 ### Added

Cargo.toml

@@ -10,7 +10,7 @@ exclude = [
 ]
 
 [workspace.package]
-version = "0.18.0"
+version = "0.19.0"
 authors = ["PulseEngine <https://github.com/pulseengine>"]
 edition = "2024"
 license = "Apache-2.0"
@@ -23,6 +23,9 @@ wasmparser = { version = "0.246", features = ["component-model"] }
 wasm-encoder = { version = "0.246", features = ["component-model"] }
 wasmprinter = "0.246"
 
+# DWARF read/write for address remapping (#143 Phase 2)
+gimli = "0.31"
+
 # CLI
 clap = { version = "4.5", features = ["derive", "cargo"] }
 

meld-cli/src/main.rs

@@ -18,7 +18,7 @@
 
 use anyhow::{Context, Result, anyhow};
 use clap::{Parser, Subcommand};
-use meld_core::{Fuser, FuserConfig, FusionStats, MemoryStrategy, OutputFormat};
+use meld_core::{DwarfHandling, Fuser, FuserConfig, FusionStats, MemoryStrategy, OutputFormat};
 use std::fs;
 use std::path::Path;
 use std::time::Instant;
@@ -80,6 +80,15 @@ enum Commands {
         #[arg(long)]
         no_component_provenance: bool,
 
+        /// DWARF debug-info handling: `strip` (default — drop all
+        /// `.debug_*`), `passthrough` (copy verbatim; addresses are
+        /// wrong against the fused code section), or `remap` (#143 —
+        /// translate code addresses to the fused code section). `remap`
+        /// currently supports a single DWARF-bearing input module and
+        /// falls back to `strip` for multi-source inputs.
+        #[arg(long, value_name = "MODE", default_value = "strip")]
+        dwarf: String,
+
         /// Preserve debug names in output
         #[arg(long)]
         preserve_names: bool,
@@ -148,6 +157,7 @@ fn main() -> Result<()> {
             stats,
             no_attestation,
             no_component_provenance,
+            dwarf,
             preserve_names,
             validate,
             component,
@@ -162,6 +172,7 @@ fn main() -> Result<()> {
                 stats,
                 no_attestation,
                 no_component_provenance,
+                dwarf,
                 preserve_names,
                 validate,
                 component,
@@ -218,6 +229,7 @@ fn fuse_command(
     show_stats: bool,
     no_attestation: bool,
     no_component_provenance: bool,
+    dwarf: String,
     preserve_names: bool,
     validate: bool,
     component: bool,
@@ -289,6 +301,18 @@ fn fuse_command(
         }
     }
 
+    let dwarf_handling = match dwarf.as_str() {
+        "strip" => DwarfHandling::Strip,
+        "passthrough" => DwarfHandling::PassThrough,
+        "remap" => DwarfHandling::Remap,
+        other => {
+            return Err(anyhow!(
+                "Invalid --dwarf mode: {}. Use 'strip', 'passthrough', or 'remap'",
+                other
+            ));
+        }
+    };
+
     let config = FuserConfig {
         memory_strategy,
         attestation: !no_attestation,
@@ -297,6 +321,7 @@ fn fuse_command(
         preserve_names,
         output_format,
         opaque_resources,
+        dwarf_handling,
         ..Default::default()
     };
 

meld-core/Cargo.toml

@@ -18,6 +18,9 @@ unexpected_cfgs = { level = "warn", check-cfg = ['cfg(kani)'] }
 wasmparser.workspace = true
 wasm-encoder.workspace = true
 
+# DWARF read/write for address remapping (#143 Phase 2 — DwarfHandling::Remap)
+gimli.workspace = true
+
 # Error handling
 anyhow.workspace = true
 thiserror.workspace = true