bpo-29347: Fix possibly dereferencing undefined pointers when creating weakref objects by zhangyangyu · Pull Request #186 · python/cpython

Splits §8 of tier7-phase3-hirbuilder-state-extraction-spec.md into
its own canonical Phase 4 spec, per supervisor 2026-04-27T08:18:04Z
authorization. Tier 7 stays focused on the original Phase 3
state-extraction question; Phase 4 sequencing/decisions/falsifiers
get a dedicated document.

Adds §5 intermediate checkpoint gates (5.A-5.E) per pythia python#186
2026-04-27 risk-flag: BLME alone (the easiest Phase 4.A pilot) won't
validate Phase 4.C class-state path or the ≥80% LOC-delete falsifier;
mid-Phase-4 close-gates close that signal gap.

Records 5 supervisor-accepted decisions (07:31:48Z) as canonical:
PARALLEL Phase 4.A entry, 3→4→5 pilot ordering, ELIMINATE
TranslationContext template emit*, DEFER compiler.cpp rewire to
Phase 4.D, full acceptance criteria including ≥10K net LOC delete.

Cross-links: post-phase3d-pure-c-roadmap.md §3 Phase 4 (parent),
tier7-phase3 + tier8 specs (siblings), BLME pilot precedent
256984d (Phase 4.A first file in flight).