task: Stability — crashes, segfaults, and memory safety during indexing (14 issues)

[DeusData]

Scope

Umbrella tracker for reported segfaults, heap corruption, hangs, and out-of-bounds reads observed during indexing. Many of these surface only at scale (large repos, deep ASTs, big all_defs[] arrays) or on specific code patterns (C++ template substitution, recovery-mode parses, deeply-nested DFS).

v0.7.0 shipped major LSP and parser-stack hardening (`1b9d9c1`, `82d5c84`) that likely already addresses several of these — reporters are encouraged to re-test on the latest release.

Additionally tracked: open PRs that target subsets of this work — PR #322 (template-call guards), PR #323 (expression eval steps), PR #339 (extract_channels growable stacks), PR #360 (C++ template call segfault).

Sub-issues

• Full-repo indexing crashes in parallel.extract on large AOSP-scale repository #141 — Full-repo indexing crashes in parallel.extract on large AOSP-scale repository
• Crash on Large Java Project: Memory Corruption Causes Database Corruption #189 — Crash on Large Java Project: Memory Corruption Causes Database Corruption
• Stuck when indexing project and laptop's fan is spinning at full speed #195 — Stuck when indexing project and laptop's fan is spinning at full speed
• SEGV in c_resolve_pending_template_calls: OOB read on NULL-terminated param_types array #215 — SEGV in c_resolve_pending_template_calls: OOB read on NULL-terminated param_types array
• Binary crash (buffer overflow) when passing unrecognised project name to MCP tools #235 — Binary crash (buffer overflow) when passing unrecognised project name to MCP tools
• segmentation fault (core dumped) #245 — segmentation fault (core dumped)
• Crash on simple C++ code snippet #312 — Crash on simple C++ code snippet
• Dump-phase crash on large TS monorepo (v0.6.1, darwin-arm64) — pipeline completes through semantic_edges then terminates abnormally before gbuf.dump #317 — Dump-phase crash on large TS monorepo (v0.6.1, darwin-arm64)
• Segmentation fault (core dumped) #336 — Segmentation fault (core dumped) — Fedora 42, icinga2
• Segmentation fault at start of lsp_cross pass on repositories with 200+ files #340 — Segmentation fault at start of lsp_cross pass on repositories with 200+ files
• SIGSEGV in lsp_cross pass when all_defs exceeds ~1189 entries (scale-dependent, not file-specific) #344 — SIGSEGV in lsp_cross pass when all_defs exceeds ~1189 entries
• Segmentation fault when running codebase-memory-mcp cli index_repository on dhw/xx_hash.h #355 — Segmentation fault on dhw/xx_hash.h
• cbm_pipeline_pass_lsp_cross deadlocks on a specific .ts file (no progress, ~3 CPU-s, indefinite) #370 — cbm_pipeline_pass_lsp_cross deadlocks on a specific .ts file
• Crash on large codebase #385 — Crash on large codebase (kicad)

Acceptance

1. Each linked issue is reproduced against the latest release; reporters confirm whether it still occurs.
2. For any remaining crash, root cause is identified and either fixed or replaced with a graceful error.
3. Soak/ASan tests cover the scale regimes that previously crashed (≥1500 files, ≥1500 defs, deep ASTs, C++ template patterns).
4. No "Segmentation fault" path in the binary — all unexpected conditions surface a logged error.

Why one task

These cluster around the parallel extract → LSP-cross → C++ template resolution paths. Treating them together avoids regressing one while fixing another, and lets us close several at once when the underlying lifecycle bug is fixed.