[JarryShaw]

• 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
• 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 — corrects a defect

---

Description of your pull request and other information

Stacked on #1998. This PR carries #1998's commit, so merge #1998 first. After that, the diff is this PR's commit alone.

Option 1 from #1999. Leaving an Extractor's with block now closes the input only if _owns_input() holds. That is the #610 rule, and _cleanup and the backstops already follow it. So:

• A stream the caller passed in stays open, whether seekable or wrapped in SeekableReader.
• A path pcapkit opened is still closed.

This reverses the choice #636 made. The 1.5.0 changelog line and the test_extraction.py case that pinned the old close are updated to match.

New ContextExitFollowsOwnership tests in test_extraction_ownership.py cover:

• a caller's seekable stream;
• a caller's non-seekable stream;
• a path closed by __exit__ alone.

With the fix reverted, the seekable case fails (2 SUBFAILED), as does test_extraction.py (1 failed).

Closes #1999

[JarryShaw]

GOOD TO GO at 883c87960, reviewing this PR's own commit on top of #1998. A Sonnet review confirmed:

• __exit__ closes the input only when _owns_input() holds, and still detaches the finaliser and closes the engine.
• A path, or a SeekableReader this class built, is still closed.
• A caller's non-seekable stream stays open, because the wrapper's stream_closing is off.
• A second close after _cleanup is a no-op.
• Reverting extraction.py fails the seekable case and the old pin in test_extraction.py. The non-seekable and path tests are regression guards.

No docs or example uses with on a caller's stream: examples/generators/dispatch.py passes a path.

[JarryShaw]

f56dc5cd6 regenerates CHANGELOG.md with util/changelog_md.py. The 1.5.0 entry's edit had left it stale, which failed Changelog drift and test_committed_changelog_is_in_step_with_the_newest_entry. Nothing else changed, so the verdict stands. tests/project/test_changelog_md.py passes locally (68 passed).

[github-actions]

Coverage: 91.54% (unit tier, Python 3.15, daa76d691, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	2	95.16%
pcapkit/const	19597	881	2618	867	92.11%
pcapkit/corekit	2540	71	808	37	96.54%
pcapkit/dumpkit	276	3	100	3	98.40%
pcapkit/foundation	2930	23	1022	25	98.79%
pcapkit/interface	138	4	52	3	96.32%
pcapkit/protocols	17468	222	4600	187	98.13%
pcapkit/toolkit	872	36	298	10	95.21%
pcapkit/utilities	523	2	158	4	99.12%
pcapkit/vendor	4431	1995	1018	167	51.73%

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

[JarryShaw]

Rebased onto main as daa76d691, now that #1998 has landed as a squash: the diff is this PR's own commit only. git range-diff shows the patch identical, so the verdict stands. Extraction and changelog tests pass.