[JarryShaw]

Please follow the guide below

• You will be asked some questions, please read them carefully and answer honestly

• Put an x into all the boxes [ ] relevant to your pull request (like that [x])

• Use Preview tab to see how your pull request will actually look like

• Searched for similar pull requests

• Followed the coding style (make pylint, make mypy, make isort)

• make test passes, and a test case covers the change. Narrow runs only, listed below.

• Added a changelog entry under docs/source/changelog/ and regenerated CHANGELOG.md, if the change is user-visible: 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?

Tick the commit type your subject line carries.

• fix — corrects a defect
• feat — adds a feature
• perf — changes performance, not behaviour
• refactor — changes neither behaviour nor performance
• test — tests only
• docs — documentation only
• ci — workflows or build tooling
• release — bumps the version or rolls up a distribution
• chore — anything else

---

Description of your pull request and other information

Closes #1706.

• Bug: StringField retried a failed decode with errors='replace' outside any try. A codec that also raises under 'replace' therefore let a UnicodeError escape unpack. That is idna on every Python, undefined, and punycode on non-ASCII octets before 3.13.
• Fix: one more fallback, to UTF-8 with 'replace', which cannot raise. UTF-8 is the fallback encoding the docstring already names, and the one the unquote branch already uses. The existing lossless check returns a DecodedString, so packing gives back the captured octets byte for byte. The errors docstring no longer says such a codec still raises.
• Trade-off: with errors='ignore', idna on b'xn--caf-dma' now reads as 'xn--caf-dma', while 'strict' still gives 'café'.
• Tests:
  • New test: test_a_codec_that_rejects_replace_does_not_raise fails on main (4 subtests on 3.14, 7 errors on 3.8) and passes on 3.8, 3.11, 3.14 and 3.15.
  • Codec sweep: a reviewer ran every codec in encodings (111 or 112) under five error schemes, over 165 byte samples, on 3.8, 3.11 and 3.14. That found 0 exceptions and 0 round-trip mismatches.

[JarryShaw]

Verdict: GOOD TO GO at 7f028562f. A Sonnet cross-review checked it; Opus wrote it.

• Codec sweep: every text codec in encodings was run under five error schemes, with unquote both on and off, over 165 byte samples, on 3.8, 3.11 and 3.14. That is about 198k cases per interpreter, with 0 exceptions and 0 round-trip mismatches.
• Before the fix: the same sweep on main raised 1,697 UnicodeErrors on 3.14, from idna and undefined. On 3.8 punycode raised as well.
• Fail before: the new test fails on main and passes on 3.8 and 3.14.
• Scope: strings.py changes in two hunks, the docstring and the fallback. No import line changes, so it does not collide with refactor(corekit)!: move chardet into corekit (#1712) #1729.

[github-actions]

Coverage: 89.35% (unit tier, Python 3.15, 7f028562f, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	3	20	3	95.16%
pcapkit/const	18860	954	2368	868	91.06%
pcapkit/corekit	2375	67	762	35	96.49%
pcapkit/dumpkit	275	3	100	3	98.40%
pcapkit/foundation	2885	144	1010	34	94.20%
pcapkit/interface	138	7	52	5	93.68%
pcapkit/protocols	17181	226	4590	190	98.07%
pcapkit/toolkit	1148	290	450	10	71.21%
pcapkit/utilities	508	4	154	4	98.79%
pcapkit/vendor	4417	2342	1010	157	43.38%

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