[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
• Added a 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 fixes

---

Description of your pull request and other information

Closes #2098. With buffer_save=True and no buffer_path, SeekableReader created a NamedTemporaryFile and later reopened it by name in read/read1/readline/peek. Windows refuses that while the file is open, so --buffer-save failed with PermissionError: [Errno 13].

• Fix: the file is created with mkstemp, and close() closes and deletes it. A weakref.finalize deletes it if the reader is collected unclosed. A caller-given buffer_path is never deleted, and a failed open releases both the descriptor and the file.
• Test: test_io_buffer_save_tempfile_2098_unit makes a second open of an open NamedTemporaryFile raise, as Windows does. Without the fix it raises PermissionError: [Errno 13]. The test also pins that the temp file is gone after close(), after collection and after an Extractor(buffer_save=True) run, and that a caller's buffer_path survives. tests/corekit/test_io.py passes.

[JarryShaw]

GOOD TO GO at a4ef1fc88 (Sonnet review).

• Test: test_the_buffer_reads_back_where_a_temporary_file_cannot_be_reopened fails with io.py reverted and passes with the fix. The other four pass on both sides; they guard against regressions, since NamedTemporaryFile used to auto-delete. Neighbouring io and extraction tests pass.
• Probes:
  • A double close() is safe.
  • close() after the file was already unlinked is safe.
  • An unclosed reader that is garbage-collected has its file removed.
  • The finalize holds only the file and the path, not the reader.
  • A caller's buffer_path is kept.
  • A failed open leaves nothing behind.
• Nit, no change: if the underlying stream's close() raises, the temp file waits for garbage collection.
• UNVERIFIED: finalisation on GraalPy and PyPy.

[JarryShaw]

Ready to merge at a4ef1fc88.

• Review: the Sonnet review is GOOD TO GO.
• CI: every blocking check passed (Required checks passed SUCCESS). The 3.8–3.10 KTLO legs carry no warnings.
• Ordering: the tests/corekit plain-unittest leg fails, with 34 failures and 31 errors, but it fails identically on plain main (aeb4aadd3). The failure set diffs as equal, every one an enum-lookup test, so it is pre-existing and not from this PR. CI has no corekit ordering leg; both points are tracked in test(corekit): the corekit plain-unittest ordering leg fails on main, and CI never runs it #2115, and a fix is in progress. The new and changed tests here pass under pytest.

This closes #2098.

[github-actions]

Coverage: 91.53% (unit tier, Python 3.15, a4ef1fc88, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	2	95.16%
pcapkit/const	19599	880	2618	866	92.12%
pcapkit/corekit	2580	75	814	37	96.46%
pcapkit/dumpkit	282	3	102	3	98.44%
pcapkit/foundation	2930	23	1022	25	98.79%
pcapkit/interface	138	4	52	3	96.32%
pcapkit/protocols	17414	219	4574	185	98.14%
pcapkit/toolkit	869	36	298	10	95.20%
pcapkit/utilities	473	7	136	5	98.03%
pcapkit/vendor	4432	1988	1018	167	51.87%

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