[JarryShaw]

Summary

• dictdumper 0.8.5 (released 2026-10-29) shipped PR Scheduled weekly dependency update for week 45 #123 (plist/json key and value escaping, fixing dictdumper PLIST writer does not escape angle brackets, so a pseudo-member makes the dump unparseable #772/DictDumper#125 and DictDumper#1152) and PR Scheduled weekly dependency update for week 01 #124 (whole-second <date> formatting, fixing fix(dumpkit): plist output is unreadable by plistlib because timestamps carry fractional seconds #1448). pyproject.toml pinned dictdumper~=0.8.0, which resolves to 0.8.5, so pcapkit.dumpkit.common's own workarounds ran on top of the upstream fix and double-escaped/mis-dated every plist/json/xml report on every fresh CI install.
• Removes escape_strings/escape_json_keys/escape_keys/escape_key() and the escaping tail of object_hook -- the writer now escapes a string or mapping key itself, for both plist/xml and json. Shrinks _append_date to just the UTC conversion dictdumper still does not do on its own, delegating the (now-fixed) whole-second formatting to super()._append_date().
• Leaves keep_none (fix(dumpkit): plist output drops every None value and None list item (dictdumper plist.py:198,228) #1257) and tree_numbers (docs(dumpkit): tree output is lossy -- decide whether to document it or fix the writer #1266) untouched -- those defects are still open upstream, confirmed by reading the installed 0.8.5 source before touching anything.
• Bumps dictdumper~=0.8.0 to dictdumper>=0.8.5 in pyproject.toml and Pipfile.
• Updates tests/dumpkit/test_common_unit.py's escaping-behaviour unit tests for the new (non-escaping) object_hook contract, and refreshes stale prose in two integration test docstrings that described the removed workarounds as still-open upstream bugs.

Closes #2442

Test plan

• Installed dictdumper==0.8.5 in a fresh venv inside this worktree; confirmed pcapkit.__file__ and dictdumper.__file__ both resolve inside the worktree before testing.
• Read the installed dictdumper/json.py (encode_string()) and dictdumper/plist.py (encode_xml(), _format_date()) to confirm which defects 0.8.5 actually fixed before deleting any workaround -- and found one it did not fix (aware-datetime-to-UTC conversion), which stays as a smaller _append_date override.
• coverage run -m pytest tests/dumpkit -- 76 passed, 1793 subtests passed (98% coverage on pcapkit/dumpkit/common.py, the three uncovered lines are pre-existing and unrelated).
• Stashed the common.py fix and confirmed the rewritten test_common_unit.py assertions fail against the pre-fix code (proving they test the right thing), then restored it.
• Ran everything that imports pcapkit.dumpkit directly (~24 files, including tests/integration/) -- all green except 4 pre-existing, unrelated failures in tests/project/test_docs_internal_refs_1940_unit.py, confirmed to fail identically on the pre-fix tree.
• mypy, pylint (project's own flag set), isort clean on pcapkit/dumpkit/common.py; no new warnings on the changed test files.

[JarryShaw]

NEEDS CHANGES (cross-review, Opus vs. authoring model).

Two required fixes, both small:

1. common.py:586 — bare (GitHub issue #1448) in a docstring needs the :issue: role; fails test_docs_issue_pr_roles_1103_unit.py::test_rendered_text_uses_roles as written.
2. The fix(dumpkit): JSON output corrupts characters above U+FFFF (dictdumper json.py:290-295) #1261 surrogate-pair fix lost coverage for JSON mapping keys: upstream's encode_string() still has the single-codepoint-escape defect for astral characters, so an emoji-bearing key now round-trips wrong. Needs a per-key surrogate-pair split on the JSON path, plus a regression test for keys (the existing test only covers values).

Non-blocking nit: common.py:334 cites "DictDumper#1152" — that's actually PyPCAPKit's own issue number; the real upstream issue is DictDumper#121.

Escaping fix, UTC conversion, and the untouched keep_none/tree_numbers workarounds all independently confirmed correct. Relayed both required fixes to the original agent.

[JarryShaw]

Both required fixes applied (new head 1cca7c8b6): the :issue: role on the docstring, and a new _append_object override for JSON mapping keys above U+FFFF (split to a UTF-16 surrogate pair before handing to upstream's encode_string), with a regression test. The optional nit (wrong DictDumper issue number) fixed too. Verified directly: diff matches the report, and tests/dumpkit/ (77 passed/1793 subtests) reproduces exactly once tested under a scratch venv with dictdumper 0.8.5 — my first local run falsely failed because the shared repo venv still has the old 0.8.4.post6. Cross-review re-dispatched on the new head.

[JarryShaw]

Review: GOOD TO GO at 1cca7c8b6 (Opus, different model from the author).

Both round-1 fixes independently re-derived as correct: the :issue: role satisfies the roles test, and the surrogate-pair key fix is sound against the full astral code-point range (0 mismatches vs. json.dumps semantics across all 1,048,576 code points) and writes no raw surrogate to the file under dictdumper 0.8.5. keep_none/tree_numbers confirmed byte-identical since main. tests/dumpkit/: 77 passed/1833 subtests. Integration: 21 passed/32 subtests.

Non-blocking: conda/requirements.txt and examples/benchmark/requirements-common.txt still pin dictdumper 0.8.4.post6 (the conda one matters for create-release.yml, which reads that file directly rather than waiting on the regen cron); top-level block names still go through unescaped encode_string; no test covers a non-str key beside an astral key.

[github-actions]

Coverage: 94.44% (unit tier, Python 3.15, 7e61bfb96, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	2	95.16%
pcapkit/const	19609	133	2620	120	98.84%
pcapkit/corekit	2605	74	818	37	96.52%
pcapkit/dumpkit	281	3	100	3	98.43%
pcapkit/foundation	3070	24	1072	26	98.79%
pcapkit/interface	138	4	52	3	96.32%
pcapkit/protocols	17593	193	4678	179	98.32%
pcapkit/toolkit	924	36	334	10	95.55%
pcapkit/utilities	502	7	144	5	98.14%
pcapkit/vendor	4449	1854	1020	167	55.11%

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

[JarryShaw]

New regression found by CI, not by either review round — the Python 3.15 leg is failing: tests/project/test_rfc_constants_one_home_1739_unit.py::TestNotReinlined::test_sweep flags pcapkit/dumpkit/common.py:328: codepoint > 65535 as an unexplained copy of a homed constant (0xFFFF/65535 coincidentally matches an existing RFC-pinned constant elsewhere, same class of false-positive LINUX_SLL's PR hit with its own 0xFFFF). Both review rounds ran narrow dumpkit/integration test selections and missed tests/project, so this slipped through GOOD TO GO. Needs a one-line NOT_A_COPY/UNRELATED registry entry in that test file, same pattern as the existing entries. Relaying the fix.

[JarryShaw]

Fixed and verified directly (new head 7e61bfb96): added the NOT_A_COPY registry entry for common.py's codepoint > 65535, following the exact pattern of the existing adjacent entry for this same file. test_sweep passes (14 passed/141 subtests). Restoring GOOD TO GO — the earlier verdict's substance still stands, this was a review-coverage gap (narrow test selection missed tests/project), not a flaw in the actual fix.