[pratyu2h]

Please follow the guide below

• Searched for similar pull requests
• Followed the coding style
• make test passes, and a test case covers the change

What is the purpose of your pull request?

• ix — corrects a defect

---

Description of your pull request and other information

Resolves #2733.

When dumping an Info or subclass instance via ields(), keys that clash with builtin names are assigned to map and their original name is appended to builtin and excluded. Previously, ields() directly checked if key not in skip, which caused keys like 'items' on Info or 'packet' on a reassembled TCP Datagram to be dropped from the dump despite being explicitly defined.

This PR checks keymap.get(key, key) not in skip where keymap = getattr(info, 'map', {}). This ensures renamed builtin keys are kept in the dumped output under their original name while still excluding regular excluded attributes. Added unit test in ests/dumpkit/test_common_unit.py.

[JarryShaw]

Thanks for picking this up. NEEDS CHANGES, and it overlaps with #2736, which was opened earlier for #2733 and has been reviewed.

• fields() is fixed, but to_dict() still records the original list in __excluded__. So dumper(info.to_dict()) and the nested branch of object_hook still drop the renamed key.
• A real reassembled datagram's packet is a deferred protocol object, not bytes. Dumped as is, it writes the protocol's internals, including an _io.BytesIO address. That makes dumps differ from run to run, which fix(dumpkit): dump a key renamed for clashing with a builtin under its own name (#2733) #2736 avoids by writing the object's .info.
• The edited fields() docstring lost the end of its sentence: "…when written." now runs into "The keys…".

Which PR to take is the maintainer's call.

[pratyu2h]

Thank you for the detailed review and guidance @JarryShaw! Closing this PR in favor of #2736.