test: capture-driven parse→rebuild and dumper round trips (#1202)

[JarryShaw]

• Searched for similar pull requests

• Followed the coding style — not run as make; pylint with the Makefile flags is clean except duplicate-code between the two modules

• make test passes, and a test case covers the change — only the two new modules (pytest and plain unittest), tests/project, test_tier_guard, test_docstring_contract

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

• test — tests only

---

Refs #1202 (partly; see below). Both modules are fixture tier, so the integration job picks them up through fixture_tier_paths(); no workflow change.

Covered

• tests/protocols/test_capture_roundtrip_runtime.py: every layer of every capture (6,6xx layers: the payload chain plus IPv6 extension headers) rebuilds from info and from info.to_dict(), Info survives to_dict/from_dict, and every capture rebuilds whole. 24 s, 223 MiB.
• tests/dumpkit/test_capture_dump_runtime.py: format='pcap' is byte-identical for all 17 PCAP inputs; json/plist/tree are deterministic, with one section per record/block counted from the file; json frames carry every Info key and the chain. 33 s, 279 MiB.

Known failures: #1446 IPv6 extension-header .data keeps the rest of the datagram that .info lacks; #1447 from_data(info.to_dict()) drops the payload and IPv6 extension headers (HTTP raises ProtocolError); #1448 plistlib cannot read any plist dump (fractional <date>), now also cited by the existing skip in test_output_formats.py.

Not covered yet: protocol edge cases, reassembly/traceflow/engine parity/registry, corekit Field/Schema symmetry, const/vendor regeneration.

[github-actions]

Coverage: 89.51% (unit tier, Python 3.14, 0bc5b3700, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	4	93.55%
pcapkit/const	18791	968	2352	869	90.90%
pcapkit/corekit	2043	89	636	25	94.77%
pcapkit/dumpkit	234	2	84	2	98.74%
pcapkit/foundation	2648	98	930	48	94.80%
pcapkit/interface	112	7	40	5	92.11%
pcapkit/protocols	16798	219	4520	183	98.08%
pcapkit/toolkit	539	74	168	3	84.58%
pcapkit/utilities	429	4	122	4	98.55%
pcapkit/vendor	4409	2342	1006	157	43.25%

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

[JarryShaw]

Verdict: GOOD TO GO at 0bc5b3700 (Sonnet cross-review; the author ran on Opus).

• The tests catch real breakage. Three deliberate mutations each turned the matching module red:
  • Ethernet make() dropping a byte: 3125 failing subtests.
  • A TCP MSS rebuilt off by one bit: 1181.
  • A JSON key reorder in dictdumper: 1616.
• The known-failure table is checked both ways. Removing an entry fails the build, and so does listing a case that doesn't exist.
• Coverage. The walk visits 6826 layers over 1616 frames in all 23 captures and stops only at NoPayload. PCAP-NG non-packet blocks are covered by the whole-file rebuild.
• Known failures reproduce on 78991b7d5. fix(ipv6): an extension header parsed with extension=True cannot rebuild from its own info #1446 rebuilds 8 of 442 octets. fix(dumpkit): plist output is unreadable by plistlib because timestamps carry fractional seconds #1448 is dictdumper/plist.py:278.
• Tier. fixture_tier_paths() lists both modules, and the tier-guard and docstring tests pass.
• Cost. About 25 s / 225 MiB and 32 s / 279 MiB. The results are deterministic across two runs, one with TZ=Asia/Tokyo.
• I re-ran both modules myself: 10 passed, 22,226 subtests.

Nit: the HTTP entry in DICT_GAPS is a different root cause from the other #1447 entries. http.py:327 _make_data accepts only Data_HTTPv1/Data_HTTPv2, not a generic data object built from a dict. I've passed this to the #1447 worker.

Merge order: the stale-entry check means that once #1446, #1447 or #1448 is fixed, its entry here must be removed in the same change. Merging this PR first lets each fix PR delete its own entries.