[JarryShaw]

• Searched for similar pull requests
• Followed the coding style
• 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?

• fix — bug fix

---

Description

Closes #2394.

• Problem: follow-up to fix: several makers reject a str enum name when their argument defaults to None #2356/fix(protocols): resolve a str enum name when its argument defaults to None (#2356) #2393, which fixed scalar _make_index arguments whose own default is None. A separate 12-site family in hip.py has the same symptom but a different shape — the str name is a loop variable over a list parameter (e.g. _make_param_dh_group_list(groups=['Group_384_bit_group'])) — so the member/int forms of the same call already worked while the str-name form raised ProtocolNotImplemented.
• Fix: each of the 12 makers gets if X_namespace is None: X_namespace = Enum_Y hoisted above its build loop, next to the existing if X is None: X = []. Chosen over inlining the conditional at the _make_index call site because at that indentation it pushed 10 of 12 lines past 120 columns.
• Two notes, not deviations: _make_param_transport_format_list's list holds parameter codes, so its namespace is Enum_Parameter — the same enum its own code argument carries. _make_param_reg_failed's scalar sibling failure_type already resolved by inference (its default is a real enum member); left alone.
• Import promotion: 5 of the 9 needed enums (Enum_Suite, Enum_NATTraversal, Enum_Registration, Enum_ESPTransformSuite, Enum_Transport) were TYPE_CHECKING-only, next to Enum_RegistrationFailure from the same pcapkit.const.hip package already imported at runtime. Promoted all 5 — pcapkit/const/hip/*.py never imports from pcapkit/protocols, so this can't create a cycle; verified with a cold import of pcapkit.protocols.internet.hip as the very first import.
• Test: tests/protocols/internet/test_hip_list_argument_str_names_2394_unit.py, 7 tests / 17 subtests — str/member/int parity for all 12 sites, a mixed list, the scalar sibling's inference, an explicit namespace override, an unknown-name raise, the 5 promoted enums, and an AST walk asserting every list-argument _make_index call in hip.py has a covering case.
• Fail-without-fix: 19 failed / 5 passed against unfixed source (the 2 passing are structural/no-regression checks, correctly fix-insensitive); all pass with the fix.

[JarryShaw]

Review: NEEDS CHANGES at 556a529ab (Sonnet) — arrived after merge, so this is now a post-merge follow-up rather than a revision request. Core fix fully verified: import-cycle claim holds (no const.hip → protocols imports, clean cold-import, no pylint R0401 cycle either direction), all 12 enum namespace choices cross-checked against the schema/reader side (136/136 str names round-trip correctly), the two flagged notes (transport_format_list's Enum_Parameter, reg_failed's unaffected scalar sibling) both confirmed correct, bug reproduces on base for all 12 sites and is fixed on head with byte-identical member/int output, 543 passed/11993 subtests, mypy/pylint unchanged, AST guard confirmed non-trivial (catches a planted 13th site).

One real issue: the new test file's prose and one assertion (assertEqual(len(found), 12)) violate the standing no-counted-numbers-in-docs ruling, which explicitly covers count pins and their tests. Filing a follow-up fix now.

Non-blocking finding from the review, worth knowing: with *_default set, origin/main silently encoded the default instead of the named value in most rows tested — not covered by this PR's scope, no test currently pins it either way.