Repository navigation
fix(registry): thirteen code-keyed registrars warn "already registered, overwriting" on re-registering the identical object #1362
Description
Describe the bug
#718 and #739 moved the registrars that store a class to an identity guard (incumbent is not x), but the registrars that store a parser (str or tuple) still test presence only, so handing back the object already stored warns about a displacement that did not happen.
| Site | Guard |
|---|---|
protocols/internet/ipv4.py:502 |
register_option |
protocols/internet/hopopt.py:387 |
register_option |
protocols/internet/ipv6_opts.py:379 |
register_option |
protocols/internet/ipv6_route.py:347 |
register_routing |
protocols/internet/hip.py:690 |
register_parameter |
protocols/transport/tcp.py:638,651 |
register_option, register_mp_option |
protocols/transport/sctp.py:645,658,671 |
register_chunk, register_parameter, register_cause |
protocols/internet/mh.py:1521,1534,1547 |
register_message, register_option, register_extension |
To Reproduce
For each class, take code, value = next(iter(cls.__option__.items())) and call cls.register_option(code, value): 13 of 13 emit RegistryWarning: ... already registered, overwriting.
Expected behavior
Silent when the incumbent is the same object (incumbent is not None and incumbent is not meth), as in #718. The auditor also listed extraction.py (4 sites) and traceflow.py:244; those already have the identity guard and are not part of this issue.
Found by the round-trip audit of pcapkit/foundation. Run /tmp/rtissues/foundation-repro.py from the repo root (it prints pcapkit.__file__ first).
Found by the foundation round-trip audit (#1202). The repro was re-checked by an independent verifier, and I re-ran it on main at c9676c563.
Activity
JarryShaw commentedon Oct 7, 2026
Blocked until the open per-protocol PRs that edit these registrar files merge: #1344 (
ipv4.py), #1342 (hip.py), #1345 (mh.py), #1348 (tcp.py) and #1310 (hopopt.py,ipv6_opts.py). A sweep across all thirteen sites now would conflict with each of them.JarryShaw commentedon Oct 7, 2026
Still blocked, narrower now: #1342 (
hip.py) and #1310 (hopopt.py,ipv6_opts.py) have merged. #1344 (ipv4.py), #1345 (mh.py) and #1348 (tcp.py) are still open.JarryShaw commentedon Oct 7, 2026
Still blocked, but only on #1367 now. #1344, #1345 and #1348 have merged. The remaining overlap is
protocols/internet/hip.py: this issue's:690site and #1367 both edit that file. The issue unblocks when #1367 merges.JarryShaw commentedon Oct 8, 2026
Still blocked, now on different PRs. #1367 has merged, so
hip.pyis clear. But this issue's sites still overlap two open PRs:protocols/internet/ipv4.py:502(register_option) is also edited by fix(ipv4)!: the TS option zeroes the timestamp slots beyond the pointer on rebuild (#1349) #1390;traceflow.py:244guard is also edited by fix(registry): foundation registrars restore the built-in a name shipped with (#1363) #1376.Blocked until #1390 and #1376 merge.
JarryShaw commentedon Oct 8, 2026
Unblocked: #1376 and #1390 have merged. A worker is starting on it. The IPv4 TS and SEC code is being worked in parallel for #1391–#1393, so this worker keeps to
register_optioninipv4.py.fix(protocols): parser registrars warn only when the entry really cha…