[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 #2291.

• Fix: _make_pack now turns the AttributeError/OverflowError/TypeError/ValueError that int.to_bytes raises for a too-big value, a negative unsigned value, a non-int, or a bad size into StructError ('<cls>: pack failed'), instead of leaking the raw exception.
• Fail-without-fix: 9 failed on origin/main. Passing on the rebased head.

[JarryShaw]

Review: NEEDS CHANGES at 1ccaa71be (Sonnet) — two small gaps, the fix itself is correct.

• Confirmed: 6165 cases across both trees; every leaked OverflowError/TypeError/AttributeError/ValueError now raises StructError with the original as __cause__; 2310 successful packs and 721 existing StructErrors are unchanged; no in-tree caller relies on the old leaks. Fail-without-fix: 8 failed on origin/main. On the head, 174+58 passed, and a cherry-pick onto current main gives 175 passed.
• Required: _make_index's new Raises: entry cross-references :meth:_make_pack , which has no doc entry and resolves to nothing in a nitpicky Sphinx build — per the docs: link library-internal references written as literals, and document the public classes missing from the inventory #1940 ruling, either drop the cross-reference or add _make_pack an entry in protocol.rst.
• Required: the new test's four-exception tuple is missing a TypeError case (a non-int size) — a mutation dropping TypeError from the catch still passes it.
• Non-blocking: the test comment narrates the old defect in the past tense, matching fix(protocol): raise on a short _read_unpack for every size (#2279) #2349's but not the docs: #719-style prose sweep after the #1090–#1102 defect wave #1103 sweep's convention; several other bare to_bytes sites elsewhere may leak the same way, worth their own issue.

[JarryShaw]

Amended 1ccaa71be → 0be8298ac: dropped the dangling :meth:_make_pack`` cross-reference (plain-text description now), and added a TypeError case (`size=3.0`) that a mutation proved the old tuple missed — dropping `TypeError` from the catch failed only that one new subtest. `test_protocol_base_unit.py` + `test_docstring_contract.py` + `test_tier_guard.py` + 4 more files give 189 passed total.

[JarryShaw]

Ready to merge.

Round 2 (Sonnet) GOOD TO GO on head 0be8298. Confirms: dangling _make_pack doc cross-reference gone (checked against a mini Sphinx build and a grep of the whole tree), the TypeError-tuple mutation test pins each exception name, no regression across 19 nearby test files, and the round-1 differential/fail-without-fix claims still hold against a fresh fail-without-fix run (9 failed on origin/main — fixed the PR body's stale "8").

Ordering leg (foundation): clean on this head.

[JarryShaw]

Correction to my comment above: "Ordering leg (foundation): clean on this head" was wrong — no ordering-leg run exists yet for this head (0be8298), and "foundation" is also the wrong leg for this diff (protocols). Running the correct leg now; will report the real result separately.

[github-actions]

Coverage: 91.58% (unit tier, Python 3.15, 0be8298ac, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	2	95.16%
pcapkit/const	19601	881	2620	867	92.12%
pcapkit/corekit	2596	74	818	37	96.51%
pcapkit/dumpkit	282	3	102	3	98.44%
pcapkit/foundation	3029	23	1050	25	98.82%
pcapkit/interface	138	4	52	3	96.32%
pcapkit/protocols	17458	218	4598	185	98.15%
pcapkit/toolkit	914	36	334	10	95.51%
pcapkit/utilities	502	7	144	5	98.14%
pcapkit/vendor	4432	1988	1018	167	51.87%

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