[JarryShaw]

• Searched for similar pull requests
• Followed the coding style (make pylint, make mypy, make isort)
• make test passes, and a test case covers the change
• Changelog entry: N/A — centralised in docs(changelog): shared 1.5.0 changelog — long-lived, merges last (#610, #616, #617, #618, #620) #657

What is the purpose of your pull request?

• chore — tooling or repository hygiene

---

Description of your pull request and other information

Plain # comments now cite issues as #NNN. The :issue:`NNN` / :pr:`NNN` roles stay only where Sphinx renders them: docstrings, #: attribute comments and .rst.

• Plain comments are converted across pcapkit/, docs/source/conf.py and tests/. The citations were found by tokenising every tracked .py file, so docstrings are left untouched. The diff changes no code tokens.
• The new tests/project/test_issue_refs_in_comments.py fails on any plain comment that uses the roles. A short allowlist covers comments that talk about the role rather than cite an issue. It fails on main and passes here.

[JarryShaw]

GOOD TO GO at d4e1d4f38. A Sonnet review found:

• Comments only: across the 29 changed files the code tokens are identical once comments are dropped, and no docstring or #: comment changed.
• Exact rewrites: all 115 comments match a mechanical :issue:``NNN``` → #NNN` rewrite.
• Guard and allowlist: the three allowlist entries are genuine meta-mentions, and the guard fails if a role is restored to a plain comment.
• Open PRs: merged with test(tiers): make ALL a curated opt-in tier (#1958) #1978, fix(protocols): handle declared lengths above sys.maxsize as overruns (#1975) #1979 and ci: pin every action to a commit SHA and let Dependabot update them (#1968) #1981, the guard still passes.