docs(changelog): link the 1.5.0 entries to the API (#1941)

[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

• Added a changelog entry — N/A — centralised in docs(changelog): shared 1.5.0 changelog — long-lived, merges last (#610, #616, #617, #618, #620) #657

• docs — documentation only

Per the ruling on #1941, names in the 1.5.0 changelog that refer to pcapkit objects are now Python-domain roles. Names 1.5.0 removed, concepts, pip extras and stdlib names stay literal. Ambiguous names carry an explicit pcapkit. path or an explicit title. Most of the .rst diff is rewrapping, so that no role splits across lines.

util/changelog_md.py renders these roles in CHANGELOG.md as code spans of the name Sphinx displays, following its rules for ~, ., ! and explicit titles, so the generated file stays readable. The only change to CHANGELOG.md is () added after callables.

The new PythonRoleTests fail on main. --check passes, and the pinned docs build reports 0 warnings. Every new role was checked against the built HTML. The older changelog pages stay literal.

Closes #1941

[JarryShaw]

GOOD TO GO at 7c111d595. A Sonnet cross-review rebuilt the docs independently: 0 warnings and no "more than one target" messages. All 754 roles render as links, and the link map is byte-identical to the worker's build. It read about 100 entries in context, including the data-versus-schema and const-versus-data pairs, and found every link landing on the object the sentence means. Removed and renamed names stay literal. With whitespace collapsed and roles stripped back to literals, the rewrapped file has the same 31,898 tokens as before. The CHANGELOG.md diff is only () additions, and PythonRoleTests fails on main. merge-tree is clean against main, #1939 and #1927.

[JarryShaw]

Im debating if we should keep hyperlinks for PR/issue references in the generated markdown changelog - i suppose they're not really necessary.

[JarryShaw]

One thing to weigh: GitHub does not auto-link #nnn in repository files. Its docs say "Autolinked references are not created in wikis or files in a repository" (autolinked references); auto-linking applies only within conversations. So if CHANGELOG.md uses plain #nnn, those references become unlinked text when the file is read on GitHub, and in the sdist copy. The rst changelog still links them on the docs site. Today the generated file has 327 issue/PR links. I'd keep them, but dropping them is a small change (util/changelog_md.py's role table at lines 196–198, then regenerate). Which do you prefer?

[JarryShaw]

okay keep them

[JarryShaw]

Keeping the links, then. CHANGELOG.md stays as generated, so this PR needs no change and stays at 7c111d595.