[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?

• docs — documentation only

---

Description of your pull request and other information

Links each rendered bare pcap module mention to the note explaining that PyPCAP and pcap-ct both install that module (#1948, option 1). The note in engines/3rdparty.rst gets a pcap-module label, and |pcap|_ in rst_epilog targets it.

An epilog target cannot reach a label on another page: an indirect target fails off its own page, and :ref: renders plain text rather than a literal. So a doctree-resolved handler in conf.py, link_pcap_module, points the link at the note's page from every other page, and warns if the label is missing.

index.rst's PCAP Wikipedia link becomes anonymous, because target names fold case and its named target clashed with pcap. Code, distribution names, engine names and the file format stay literal.

The hook leaves singlehtml's single-page anchor alone and writes the plain literal where a builder has no page for the note (latex without it). tests/project/test_docs_pcap_module_link_1948_unit.py, including real singlehtml and latex builds, fails without the change. The docs build on the pinned versions has no warnings.

Closes #1948

[JarryShaw]

NEEDS CHANGES at bc9c279d8, one wording nit. A Sonnet delta review checked both builder rules against the Sphinx 9.1 source. Only singlehtml returns a # page URI, so its bare anchor is left alone. latex and texinfo raise NoUri for a page outside the document set, so the link becomes its literal. text and man build with 0 warnings. The new builder tests fail without the fix, and the html build has 0 warnings. The one change, in a follow-up commit: a test docstring says the build "used to die", which is timed wording.

[JarryShaw]

GOOD TO GO at 559956780. A Sonnet delta review confirmed the new commit rewords only the latex test's docstring and the KEPT comment, with no assertion changed. The new wording matches what the test asserts, and no history wording is left in the lines the PR adds. The test file passes in both the repo venv and the pinned docs venv (17 passed).