100%

Section 03: Code Journeys — Expanded Coverage

Status: Not Started Goal: Add 3 new code journeys (J14-J16) covering the patterns that were missing from the original 13 — specifically heap-typed loop reassignment, nested RC structures, and COW mutation patterns. All must score 10/10 and be leak-free.

Context: None of the original 13 journeys exercise heap-typed variable reassignment inside a loop. J7 (loops) uses int accumulators; J9 (strings) uses linear concat without loops; J10 (lists) iterates without reassignment. This gap let the FatValue PrimOp crash slip through to near-merge. J14-J17 (fat pointer journeys, added 2026-03-16) test fat pointer codegen basics but NOT these specific gap patterns. These new journeys (J18-J20) close the gap.

Depends on: Section 01 (leak detection infrastructure) and Section 02 (leak fixes must be complete so journeys don’t crash or leak).

03.1 J18 — String Builder (Loop Heap Reassignment)

File(s): plans/code-journeys/18-string-builder.ori, plans/code-journeys/18-string-builder-results.md

Journey theme: “I am a string builder.” Exercises the exact pattern that caused the crash — string concatenation in a loop with heap promotion from SSO.

  • Create 18-string-builder.ori with:
    • @build_repeated(n: int, c: str) -> str — loop appending c to a string n times
    • @build_sequence(n: int) -> str — loop appending str(i) for each i in 0..n
    • @build_with_separator(items: [str], sep: str) -> str — loop joining strings with separator
    • @main — combines results, returns checkable integer
  • Verify interpreter produces correct result
  • Verify AOT produces correct result
  • Run diagnostics/dual-exec-verify.sh on J18 — interpreter and AOT produce identical output
  • Run ORI_CHECK_LEAKS=1 — zero leaks
  • Run valgrind — zero errors, zero bytes at exit
  • Run /code-journey skill (via Skill tool) on the journey file — full pipeline: traces, deep scrutiny, scoring, results file
  • Create 18-string-builder-results.md with full scoring breakdown
  • Target: 10/10 overall (if below 10/10, fix the underlying issue before proceeding to Section 04)

03.2 J19 — RC Lifecycle (Nested Heap Structures)

File(s): plans/code-journeys/19-rc-lifecycle.ori, plans/code-journeys/19-rc-lifecycle-results.md

Journey theme: “I am a lifecycle.” Exercises structs containing heap fields, passing them to functions, extracting fields, and letting them go out of scope.

  • Create 19-rc-lifecycle.ori with:
    • type Container = { items: [int], name: str } — struct with two heap fields
    • @make_container(n: int) -> Container — construct with heap fields
    • @extract_and_use(c: Container) -> int — project fields, use, let go
    • @pass_through(c: Container) -> Container — identity (tests ownership transfer)
    • @nested_containers() -> int — struct containing another struct with RC fields (exercises recursive aggregate drop)
    • @main — exercises all patterns, returns checkable integer
  • Verify interpreter produces correct result — exit 51 ✓ (2026-03-20)
  • Verify AOT produces correct result — exit 51 ✓ (2026-03-20). Required fix: defined-dead project-borrowed vars getting spurious RcDec (emit_defined_dead in walk_dec.rs). Fixed by skipping all borrowed defs in defined-dead context.
  • Run diagnostics/dual-exec-verify.sh on J19 — behavioral equivalence — DUAL-EXECUTION: ALL VERIFIED (2026-03-20)
  • Run ORI_CHECK_LEAKS=1 — zero leaks — clean exit, no RC tracking output (2026-03-20)
  • Run valgrind — zero errors — 0 errors from 0 contexts, 16 allocs, 16 frees, 0 bytes in use at exit (2026-03-20)
  • Run /code-journey skill (via Skill tool) on the journey file — full pipeline: traces, deep scrutiny, scoring, results file (2026-03-20)
  • Create 19-rc-lifecycle-results.md with full scoring breakdown — 10.0/10, all 7 functions OPTIMAL, zero ARC violations, 100% attribute compliance (2026-03-20)

03.3 J20 — COW Patterns (Shared vs Unique Mutation)

File(s): plans/code-journeys/20-cow-patterns.ori, plans/code-journeys/20-cow-patterns-results.md

Journey theme: “I am copy-on-write.” Exercises the COW runtime — unique owner mutations (in-place), shared owner mutations (copy), and the transition between them.

  • Create 20-cow-patterns.ori with:
    • @unique_append() -> int — unique string concat (should mutate in place)
    • @shared_fork() -> int — share a string, mutate one copy, verify original unchanged
    • @list_cow_loop() -> int — list push in a loop with COW fast path
    • @slice_cow() -> int — create slice from list, verify original unaffected, drop both cleanly
    • @main — combines results (returns 105)
  • Verify interpreter produces correct result — exit 105 (2026-03-20)
  • Verify AOT produces correct result — exit 105, matches eval (2026-03-20)
  • Run diagnostics/dual-exec-verify.sh on J20 — behavioral equivalence — DUAL-EXECUTION: ALL VERIFIED (2026-03-20)
  • Run ORI_CHECK_LEAKS=1 — zero leaks — clean exit, no RC tracking output (2026-03-20)
  • Run valgrind — zero errors — 0 errors from 0 contexts, 9 allocs, 9 frees, 0 bytes in use at exit (2026-03-20)
  • Run /code-journey skill (via Skill tool) on the journey file — full pipeline: traces, deep scrutiny, scoring, results file (2026-03-20)
  • Create 20-cow-patterns-results.md with full scoring breakdown — 10.0/10, all 5 functions OPTIMAL, zero defects, 100% attribute compliance (2026-03-20)

03.R Third Party Review Findings

  • None.

03.N Completion Checklist

  • J18, J19, J20 source files created and committed (2026-03-20)
  • All 3 new journeys produce correct output on both eval and AOT — J18=67, J19=51, J20=105 (2026-03-20)
  • All 3 new journeys pass diagnostics/dual-exec-verify.sh (behavioral equivalence between interpreter and AOT) (2026-03-20)
  • All 3 new journeys have zero leaks (ORI_CHECK_LEAKS=1) (2026-03-20)
  • All 3 new journeys have zero valgrind errors (2026-03-20)
  • All 3 new journeys score 10/10 — J18=10.0, J19=10.0, J20=10.0 (2026-03-20)
  • Results files created in plans/code-journeys/ (2026-03-20)
  • plans/code-journeys/overview.md updated with J18-J20 (2026-03-20)
  • Original 17 journeys still score 10/10 (no regression) — all 20 journeys at 10.0/10 (2026-03-20)

Exit Criteria: 20 total code journeys (17 existing + 3 new), all scoring 10/10, all leak-free, all valgrind-clean. The new journeys specifically cover heap loop reassignment, nested RC structures, and COW patterns — the three gap areas.