Demo — maze generator + solver

• Type: feature
• Status: done
• Owner: Codex (Track B)
• Opened: 2026-06-19
• Closed: 2026-06-22
• Relation: demo-class candidate from idea-demo-app-candidates (top-10). Pressures the set lane (visited-set) — see feature-language-gaps-from-demos Gap 1. ASCII-over-serial, like the sudoku game. Platonic source.

Goal

Generate a maze from a seed (recursive-backtracker / Prim), then solve it (BFS / DFS / A*), rendering both as ASCII. Seed → reproducible maze + path.

Surface / shape

• grid of cells with wall bitflags (set of directions, or a bitmask)
• generator: seeded PRNG (app-local, as in the sudoku game) carving passages
• solver: BFS/A* over the grid — queue / priority collection
• ASCII renderer (managed strings)

Coverage

2-D arrays · sets (visited cells / wall directions — the runtime set lane) · collections (frontier queue / priority) · recursion or explicit stack · seeded integer PRNG · managed strings (render). Integer-deterministic.

Acceptance / oracle

• Fixed seed → fixed maze layout + fixed solution path length, byte-identical across targets.
• Demo: examples/maze/ prints the maze and the solved path; optional interactive "walk it" mode over serial.

Constraints

Platonic source; no compiler changes; ESP32-fit (tune grid size per target). No self-host / cross regression.

Log

• 2026-06-19 — Opened in the demo-ticket organization pass.
• 2026-06-22 — Re-verified by Codex on Track B pinned stable v32: stable_linux_amd64/default/pinned examples/maze/maze.pas /tmp/maze_pxx && /tmp/maze_pxx builds and runs the solver successfully. The demo renders the solved fixed-seed maze and reports seed 12345 size 12x8 path cells = 43. No code change or new implementation commit was needed; this ticket was stale after later compiler/runtime fixes.

Status 2026-06-19 (track B, v10)

Compiles on v10 and Generate + Render work (prints the maze). Solve (BFS) segfaults at runtime — isolated: replacing the Solve call with a constant renders fine; a minimal function with the same large 2-D static-array locals does NOT crash, so it's specific to Solve's BFS/path logic (a subtle OOB on my side, or a context-sensitive codegen issue — no clean standalone repro yet). In demos it shows compile-OK; runtime solve is the open item. Revisit with gdb / step-through once there's time; not a clean track-A ticket yet.