[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. Only the touched module was run, on 3.8, 3.9, 3.10 and 3.14. The full suite is left to CI.

• 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 of your pull request and other information

_Registries.__or__/__ror__ called dict.__or__/dict.__ror__ without a check. Dict union is new in 3.9 (PEP 584). On 3.8, any | or |= with AppType.__registries__ imported TCP and UDP and then raised AttributeError, where a plain dict raises TypeError.

• Fix: both methods are now defined only if sys.version_info >= (3, 9), a version gate in the style of pcapkit/utilities/compat.py. __dir__ leaves out the new sys import, as it already does importlib. On 3.8, all four forms (r | {}, {} | r, r |= {}, {} |= r) now raise dict's TypeError, naming '_Registries', and import nothing. A 26-case probe of union forms (operator, dunder and hasattr calls) gives the same output before and after on 3.9, 3.10 and 3.14. |= keeps dict's own in-place update, which reads no value, so the registry stays pending.
• Tests: the or/ror views run from 3.9 only, and an assertion pins which views ran. The new test_union_is_dict_s_on_every_python compares the four forms with dict's behaviour on the running interpreter.
• Fail-before: on origin/main under 3.8.20, the two tests named in the issue fail with AttributeError (2 failed), and the new test errors in all four subtests. On 3.9.25, main passes 14 of 14.
• After: 15 passed on each of 3.8.20, 3.9.25, 3.10.21 and 3.14.8. Line and branch coverage of apptype/__init__.py is 100% on 3.14. pylint, mypy and isort report nothing new.
• Other overrides: every other method the registry overrides already exists on 3.8's dict. The 3.10+ keys().mapping reads go through the registry's own methods.

Closes #1694

[JarryShaw]

Verdict: GOOD TO GO at 6be6781e4. A Sonnet cross-review reproduced #1694 on 3.8 at base: AttributeError: type object 'dict' has no attribute '__or__'. On this head the module passes 15 tests on 3.8, 3.9, 3.10 and 3.14. It ran a 26-case probe of every union form in a fresh process per case. The output is byte-identical between base and head on 3.9, 3.10 and 3.14. On 3.8, union now raises dict's own TypeError.

[github-actions]

Coverage: 89.47% (unit tier, Python 3.15, 1691825be, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	3	20	3	95.16%
pcapkit/const	18860	954	2368	868	91.06%
pcapkit/corekit	2372	67	762	35	96.49%
pcapkit/dumpkit	275	3	100	3	98.40%
pcapkit/foundation	3035	151	1076	34	94.28%
pcapkit/interface	138	7	52	5	93.68%
pcapkit/protocols	17166	226	4580	190	98.07%
pcapkit/toolkit	1437	279	592	5	78.12%
pcapkit/utilities	494	4	146	4	98.75%
pcapkit/vendor	4417	2342	1010	157	43.38%

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

[JarryShaw]

prefer to use version gates, instead of hasattr checks

[JarryShaw]

Recorded: a sys.version_info gate instead of hasattr(dict, '__or__'). A worker is on round 2; the verdict resets on the new head. The same applies to #1701's round 2, which uses sys.version_info blocks in compat.py in place of its getattr chain, and to #1709.

[JarryShaw]

Verdict: GOOD TO GO at 1691825be. A Sonnet cross-review checked round 2. The hasattr(dict, '__or__') probe is now a sys.version_info >= (3, 9) gate in compat.py's tuple style. sys is hidden from the module's eager dir(), and branch coverage on 3.14 is 100% with the existing pragma. The reviewer ran its own 30 cases in a fresh process each, on 3.8, 3.9, 3.10 and 3.14. The 26 round-1 cases are byte-identical to round 1. On 3.9, 3.10 and 3.14 the results match main; on 3.8, every union form raises dict's own TypeError and imports nothing. The module passes on all four interpreters.