[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

GraalPy's struct packs a bytearray for the c format, but CPython's raises struct.error. Field.pack already converts a bytearray to bytes for s and p (#2004); it now does the same for c, so a bytearray packs identically on every engine. A memoryview is not converted, because CPython refuses one for c, s and p alike; it is passed straight through to struct. No library field uses c.

What the tests check:

• test_bytearray_for_the_c_format_packs_as_bytes: a bytearray, or a subclass of one, packs the same as b'a', both under the real struct and under a stand-in that accepts buffers the way GraalPy's does. With main's field.py, it fails 4 subtests.
• test_memoryview_for_the_c_format_is_left_to_struct: Field.pack hands the memoryview to struct unchanged.

GraalPy is not installed locally, so this was not run there. Only narrow selections were run, not make test: test_fields_bytearray_pack_2004_unit.py and test_fields_field.py pass.

Closes #2015

[JarryShaw]

GOOD TO GO at 4c8ce4fb3.

A Sonnet cross-review checked every claim it set out to:

• No library field uses a c template, and template[-1:] handles a prefixed >c.
• StructError subclasses struct.error, so existing except struct.error handlers still catch it.
• With main's field.py, the test file fails 6 subtests; with the fix, it gives 47 passed.
• bytes and its subclasses still pack.

The behaviour has not been checked on GraalPy itself: the test stands in for GraalPy's struct, so the graalpy-25.4 leg will be the first real check.

[github-actions]

Coverage: 91.55% (unit tier, Python 3.15, f9fe66a76, 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	2542	71	810	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	17501	222	4612	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]

cant we just convert like the line before?

[JarryShaw]

Yes. I'll change the fix so that a bytearray for the c format is converted to bytes, the same as for s and p, instead of being refused. It then packs the same on every engine. The test will pin that a bytearray packs to the same bytes as the equivalent bytes value, both under a strict struct and under one that accepts buffers, the way GraalPy's does. I'll check memoryview the same way and report what I find. Until the new head is reviewed, this PR is back at review: needs-changes.

[JarryShaw]

GOOD TO GO at f9fe66a76.

A Sonnet re-review confirmed:

• The StructError refusal, its import and its Raises: entry are all gone.
• A bytearray, or a subclass of one, for c packs as b'a'. A length that isn't 1 still raises struct.error.
• With main's field.py, the new test fails 4 subtests. On the head, 5 passed and test_fields_field.py gave 42 passed.
• The memoryview test checks only that the value is passed through, so it can't fail on GraalPy for an engine reason.

GraalPy itself was not run.