[JarryShaw]

Please follow the guide below

• Searched for similar pull requests
• Followed the coding style (make pylint, make mypy, make isort) — ran pylint (Makefile flags), mypy and isort on the two touched files directly, not via make; no new findings beyond the existing protected-access pattern
• make test passes, and a test case covers the change — the modules listed below only, not the full suite
• Added a changelog entry — N/A, added centrally after the wave

What is the purpose of your pull request?

• fix — corrects a defect

---

Description of your pull request and other information

Closes #1154.

• http.HTTP._make_data looked up a version key that neither versioned data class has, so it always raised invalid HTTP version: 0. It now dispatches on the data class and returns the version that make needs.
• httpv2.HTTP._make_data returned length, which ProtocolBase.__init__ takes as the parse length (payload only), so HTTP/2 from_data failed even directly (packet (6 octet(s)) shorter than the 9-octet frame header). Dropped; make computes the Length.
• Two test_http_unit cases drove _make_data with a fake version key (and pinned length); they now use parsed data.

Probe, HTTP.from_data(frame['HTTP'].info) over the captures: before, 222/4/5 × invalid HTTP version: 0 (http/options-transport/test.pcap). After, all 231 rebuild byte-exact with equal info.

New test_http_from_data_1154_unit: 4 passed, 31 subtests. Without both fixes: 32 failed, 3 passed. Without only the httpv2 fix: 28 failed. Also passing: test_http_unit, test_http_guess_http2_1141_unit, test_http_runtime, four test_httpv2_*, test_construction_keyword_check_unit, test_option_roundtrip_unit, test_tcp_http_dispatch_unit, test_frame_from_data_runtime, tests/project (389 passed, 1 skipped).

[JarryShaw]

Cross-review verdict on cb7afe6ba: GOOD TO GO (ran on Sonnet; author Opus)

• Both root causes confirmed on main. In http.py, _make_data dispatched on data.get('version', 0), but neither data class has that field. In httpv2.py, _make_data returned the payload-only length, and ProtocolBase.from_data (protocol.py:798) passes that length to unpack as the parse length. I re-read both myself.
• Byte-exact: 231/231 HTTP frames rebuild with equal bytes and info. With test(generators): HTTP/2 sample frames are all written on stream 0 #1161 applied the count is 237/237.
• Edited tests: the fake version key pinned a defect. The only caller of _make_data is from_data. The old message invalid HTTP version: N is still raised from read/make (:193, :230), and nothing pins the _make_data text.
• Fails without the fix: 32 failed with both sources reverted, and 28 with only httpv2.py reverted.
• Follow-ups, not blocking:
  • A frame parsed after an HTTP/2 connection preface rebuilds without the preface (39 → 15 octets), because info does not carry it.
  • link/l2tpv2.py:367 _make_data also returns length. That is likely the same class of bug, but it is unverified on a real capture.

[github-actions]

Coverage: 88.87% (unit tier, Python 3.14, cb7afe6ba, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	4	93.55%
pcapkit/const	18757	1035	2342	846	90.23%
pcapkit/corekit	1878	91	580	22	94.34%
pcapkit/dumpkit	142	0	42	0	100.00%
pcapkit/foundation	2422	104	842	40	94.24%
pcapkit/interface	112	7	40	5	92.11%
pcapkit/protocols	15690	188	3960	163	98.18%
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.