refactor!: drop the surviving re-exports of moved names (#1744)

[JarryShaw]

Please follow the guide below

• You will be asked some questions, please read them carefully and answer honestly

• Put an x into all the boxes [ ] relevant to your pull request (like that [x])

• Use Preview tab to see how your pull request will actually look like

• 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. Narrow runs only, listed below.

• Added a changelog entry under docs/source/changelog/ and regenerated CHANGELOG.md, if the change is user-visible: 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?

Tick the commit type your subject line carries.

• fix — corrects a defect
• feat — adds a feature
• perf — changes performance, not behaviour
• refactor — changes neither behaviour nor performance
• test — tests only
• docs — documentation only
• ci — workflows or build tooling
• release — bumps the version or rolls up a distribution
• chore — anything else

---

Description

Closes #1744. It applies the owner's rulings on #1715, #1103 and #1744 (D1–D5): a moved name keeps no explicit re-export at its old path. An explicit re-export here means an alias assignment, a name in a non-home module's __all__, or prose saying the name is importable there. Ordinary imports that a module uses itself stay.

Breaking. These old paths no longer resolve:

• protocols.misc.pcapng.TLSKeyLabel and corekit.fields.field.NoValueType shipped in 1.4.x, so both now carry versionremoved:: 1.5.0, in the docstring and on the docs page.
• The beta-only names are deleted with no directive:
  • NO_VALUE, NO_DEFAULT/NoDefaultType, NULL/NullType and ABSENT/AbsentType, from their non-home __all__ or bindings;
  • the _get_null pickle shim (D2);
  • DeferredPacket in the reassembly and traceflow data modules;
  • the ngap aliases ProcedureCode and ProtocolIE.
• Registry NULL gets versionchanged:: 1.5.0, because in 1.4.x it was '(null)' (D4).

Every reader now imports from the new home. The vendor ftp and apptype templates, and the two const modules they generate, change identically. The sentinel docs no longer describe re-exporting the sentinels from each module.

Guard, tests/project/test_no_moved_name_re_exports_1744_unit.py. It fails on any of these for a removed (module, name) pair:

• the name in that module's __all__, or still bound there;
• an ImportFrom reader, or an attribute read through the old module;
• an auto* directive naming the old path;
• prose under pcapkit/ or docs/ claiming the name is importable there.

It also checks that both versionremoved blocks are present.

Tests, narrow runs on Python 3.14: the guard plus tests/corekit, the tests/foundation reassembly, traceflow and registry suites, pcapng, ngap, the const modules and the tests/project gates. All pass, 1,600+ tests.

[JarryShaw]

Verdict: GOOD TO GO at 79c92af89. A Sonnet cross-review checked two rounds (author: Opus).

• Round 1 found one blocker: the DeferredPacket docstring (corekit/packet.py) still said the class was importable from its three old paths, and the guard had no check for that kind of prose. Everything else held when the reviewer re-derived it independently:
  • the re-export scans and the readers;
  • the version directives, against tags v1.4.0…v1.4.1.post2: only TLSKeyLabel and NoValueType shipped;
  • the vendor-template text;
  • 9 of 10 mutants.
• Round 2:
  • the sentence is gone;
  • the guard now also catches attribute reads through an old module and re-export prose;
  • the branch is rebased onto 3946bf527 with an identical patch (git range-diff);
  • 11 mutants all fail the guard, including the reviewer's own (import pcapkit; pcapkit.corekit.module.NULL, a claim wrapped over three lines, readers under examples/ and util/).
• Non-blocking: partial-package aliases (import pcapkit.corekit as ck; ck.module.NULL), importlib lookups and rarer wordings ("can still be imported from") still pass the guard. No such case exists in the tree.
• My own check: the guard plus test_unit_tests_queue, test_public_api and both sentinel test files give 82 passed and 1055 subtests, with pcapkit.__file__ printed inside the PR's worktree.

[github-actions]

Coverage: 89.59% (unit tier, Python 3.15, 79c92af89, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	3	20	3	95.16%
pcapkit/const	18862	954	2368	868	91.06%
pcapkit/corekit	2461	69	780	35	96.54%
pcapkit/dumpkit	276	3	100	3	98.40%
pcapkit/foundation	2915	164	1016	34	93.59%
pcapkit/interface	138	7	52	5	93.68%
pcapkit/protocols	17184	224	4574	189	98.08%
pcapkit/toolkit	872	177	298	8	77.18%
pcapkit/utilities	512	2	158	4	99.10%
pcapkit/vendor	4417	2342	1010	157	43.38%

Per-file detail: the coverage-html artifact of this run.