[JarryShaw]

• Searched for similar pull requests

• Followed the coding style — isort -l100 -ppcapkit clean; pylint/mypy on protocol.py report nothing new on the changed lines (the 117-col raise matches the pre-existing one)

• make test passes, and a test case covers the change — only the modules listed below

• Changelog entry — N/A, added centrally

• fix — corrects a defect

---

Root cause. #1344 pads the IPv4 option area once, at the end. So [b'\x02', b'\x94\x04\x00\x00'] now packs as 02 94 …: option 2 claims length 148. The schema accepts that, and _read_ipv4_options rejects it at ipv4.py:659 (IPv4: invalid format). That check runs in read(), outside the try that #1352 put around the schema re-parse only (protocol.py:737-742). Before #1344, each item got its own padding, 02 01 01 00. That gave a negative data length, which raised inside the try.

Which side is right. #1352's side. The input is still malformed, so the test stays as it is. unpack() now also wraps read() when the made header holds raw items. The wrapped message adds the original error's text after a colon (… given to make(): IPv4: invalid format), so a regex on the reader's message still matches.

Other protocols. The same escape happened in TCP, HOPOPT, IPv6-Opts, MH, SCTP and HIP. A fuzz of 400 raw items per protocol on main let 517 bare read() errors out: IPv4 375, MH 70, TCP 36, HOPOPT 15, IPv6-Opts 13, SCTP 8. HIP's escape is the ESP_INFO case in the tests. With this fix, all of them are wrapped. Errors raised inside make() itself are unchanged (TCP _make_tcp_options enum lookup, IPv6-Route type check).

Tests. New module tests/protocols/test_make_raw_item_read_unit.py, with 8 cases that read() rejects, plus a well-formed guard. Revert proof: with git show origin/main:pcapkit/protocols/protocol.py written over the file, all 8 subtests fail, plus the original test_unparsable_raw_item_raises_protocol_error. With the fix, these pass:

• test_protocol_make_unit.py: 12 passed
• new module: 2 passed, 8 subtests
• leg protocols/internet: 546 tests, OK
• leg protocols --exclude protocols/internet: 968 tests, OK
• tests/project: 389 passed, 1 skipped

Contract note: the ProtocolError text for a malformed raw make() item now ends with the original message.

Closes #1394

[github-actions]

Coverage: 87.50% (unit tier, Python 3.14, 0c22e7d27, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	4	93.55%
pcapkit/const	18777	1032	2348	843	90.27%
pcapkit/corekit	2018	129	628	42	92.25%
pcapkit/dumpkit	189	1	68	1	99.22%
pcapkit/foundation	2443	347	852	103	82.40%
pcapkit/interface	112	7	40	5	92.11%
pcapkit/protocols	16233	398	4276	261	96.57%
pcapkit/toolkit	539	74	168	3	84.58%
pcapkit/utilities	429	47	122	2	85.30%
pcapkit/vendor	4409	2359	1006	158	42.84%

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

[JarryShaw]

Cross-review verdict on 0c22e7d27: GOOD TO GO (Sonnet reviewed; author Opus). This unblocks main (#1394).

• Scope of the wrap. It applies only in the _make_raw and _holds_raw(...) branch, so make() without raw items is unchanged. It catches (BaseError, struct.error) only. BaseError is an Exception subclass, so KeyboardInterrupt, MemoryError and RecursionError propagate. raise … from error keeps __cause__.
• The message change is needed. main's own test at test_protocol_make_unit.py:162 expects [ParamNo 65] invalid format in the message. I checked that line myself.
• Tests. test_protocol_make_unit plus the new module give 14/0. The revert proof gives 9 failed: the original test and all 8 per-protocol cases (IPv4 overrun and short, TCP, HOPOPT, IPv6-Opts, MH, SCTP, HIP).
• CI. Green at this head (38/0), and the branch is up to date with main.

Nit: a ProtocolError raised by read() gets wrapped again, so the message nests. __cause__ is still kept.

An earlier review run went past its time budget and was stopped before reporting; this verdict comes from a narrower re-run.