[JarryShaw]

• Searched for similar pull requests
• Followed the coding style (make pylint, make mypy, make isort) — ran pylint (Makefile flags), mypy and isort on protocol.py only: no new findings in the changed region
• make test passes, and a test case covers the change — the legs listed below, not the full make test
• Added a changelog entry — N/A, added centrally

What is the purpose of your pull request?

• fix — corrects a defect

---

Description of your pull request and other information

Closes #1210

_make_index now infers the namespace when it gets a str and no namespace. It finds the calling make() / _make_* method on the class and picks the arguments whose value equals the name (==, not identity), whose <arg>_namespace is None, and whose default is an enum member. It then uses that default's enum class. When several arguments share the name, e.g. IPv4(tos_thr='HIGH', tos_rel='HIGH'), every candidate enum must have it with the same value. Otherwise it raises ProtocolNotImplemented: ... ambiguous member name ...; pass a_namespace= or b_namespace= explicitly, e.g. IPv4(tos_del='LOW', tos_thr='LOW'), since LOW is not a ToSThroughput member. No caller changed. int, enum members and explicit namespaces never reach the new code. 94 arguments across 20 modules are now inferable.

Not inferable, so these keep the old behaviour: HTTP.make(status=), HOPOPT/IPv6_Opts _make_opt_smf_dpd(tid_type=) and _make_opt_mpl(seed_type=), all of which default to None. Also the list arguments of 12 HIP _make_param_* methods (dh_group_list, hip_transform, hip_cipher, nat_traversal_mode, hit_suite_list, transport_format_list, esp_transform, hip_transport_mode, reg_info/request/response/failed). Their items are passed through a loop variable, not through a parameter with an enum default.

Before / after: Ethernet(type='Internet_Protocol_version_6') raised ProtocolNotImplemented. It now gives …86dd, the same bytes as the enum member. L2TPv2(type='Control') now gives c002000800000000.

Tests: new tests/protocols/test_make_index_str_unit.py covers link (Ethernet, L2TPv2), internet (IPv4), transport (TCP option), application (OSPF) and misc (PCAP header). Each case checks str == member == int, plus the non-default member. It also covers both ToS collisions and an ambiguous case, both a constructed one and the real LOW one. Against main it fails 8; against the first revision of this PR it fails the 2 collision tests; with the fix, 7 passed and 6 subtests passed.

• run_unittest_leg.py protocols --exclude protocols/internet: 857 tests, 0 failures, 0 errors
• run_unittest_leg.py protocols/internet: 503 tests, 0 failures, 0 errors
• pytest tests/project (first revision): 389 passed, 1 skipped

[github-actions]

Coverage: 88.92% (unit tier, Python 3.14, d7c505ffd, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	4	93.55%
pcapkit/const	18757	1034	2342	845	90.24%
pcapkit/corekit	1878	91	580	22	94.34%
pcapkit/dumpkit	142	0	42	0	100.00%
pcapkit/foundation	2442	102	852	39	94.38%
pcapkit/interface	112	7	40	5	92.11%
pcapkit/protocols	15819	191	4064	166	98.17%
pcapkit/toolkit	487	65	144	1	85.42%
pcapkit/utilities	429	4	122	4	98.55%
pcapkit/vendor	4409	2359	1006	158	42.84%

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

[JarryShaw]

Cross-review verdict on bb5c9a881: NEEDS CHANGES (Sonnet reviewed; the author ran on Opus).

• Must fix: _make_index_namespace matches the caller's argument by identity (values.get(arg) is name). When two enum arguments receive the same literal, both match. Inference then gives up and raises ProtocolNotImplemented. Examples: IPv4(tos_del='NORMAL', tos_thr='NORMAL') and IPv4(tos_thr='HIGH', tos_rel='HIGH'). Whether it fails depends on string interning: 'NOR'+'MAL' works. main also raises here, so this is not a regression, but the gap is neither tested nor documented.
• Confirmed: across 376 classes, the default make() output is byte-identical to main. So is member/int output for 6 members of every enum argument. 111/111 string names build the same bytes as their members. Invalid names still raise or fall back as before. There is no measurable slowdown on the non-str path. Both test legs are clean (498/0 and 855/0).
• Count: 21 modules call _make_index, not 22.

[JarryShaw]

Cross-review verdict on d7c505ffd: GOOD TO GO (Sonnet, round 2; author Opus).

• Fixed: IPv4(tos_del='NORMAL', tos_thr='NORMAL') now builds the same bytes as the enum form, and so does a non-interned 'NOR'+'MAL'. tos_thr/tos_rel='HIGH' matches too.
• No other output changed: default make() output is byte-identical to main across every protocol class.
• Tests: the new tests fail on main. The legs give 503/0 and 857/0.
• Follow-up, not a regression (filed separately): the strict rule over-raises. tos_thr='LOW', tos_thr_default=1 falls back as it does on main, but adding tos_del='LOW' makes the same call raise "ambiguous". No two enum arguments currently share a name with different values, so the value-conflict branch never fires. The fix is to resolve each _make_index call against its own argument, which all 111 call sites name plainly. The hint in the error message also points at tos_thr_namespace=, which could not help in that case.