Three loose ends from #509 (PR #509, merged as 27bb315d5). None is a behaviour regression — the refactor is byte-identical, verified — but each undercuts a claim the refactor itself makes. The CGA length defect it also exposed is filed separately as #512.
1. mh.py:2853 still open-codes the arithmetic the helper was created to collect
pcapkit/protocols/internet/mh.py:2730-2732 states the intent:
collecting the read-side half of it into one helper is so a future fix to this arithmetic only has to happen once
But _read_opt_pad at :2853 still reads:
if code == Enum_Option.Pad1:
size = 1
else:
size = clen + 2where the equivalent site in pcapkit/protocols/internet/hopopt.py:603 was converted:
_size = self._hopopt_option_length(schema.len)
ipv6_opts.py:612-615 was converted too. Only the mh.py branch was left behind, so the arithmetic still exists in two places in the file that introduced the helper.
2. A cross-reference that points at nothing
mh.py:2727-2730 refers the reader to "the surviving explanation of that fix … below". That Note: exists only in hopopt.py:579-588 and ipv6_opts.py:590-599; mh.py's _read_opt_pad has no Note: at all. The paragraph was copied and its "below" did not travel with it.
3. ipv6_opts.py:467 cites the wrong RFC section
Both helpers cite :rfc:8200#section-4.3``:
hopopt.py:456 — correct, §4.3 is the Hop-by-Hop Options Headeripv6_opts.py:467 — wrong, IPv6-Opts is Destination Options, §4.6
The same file already uses #section-4.6 correctly for the same kind of claim at :1253. Introduced by #509's own new docstring, copied verbatim from hopopt.py. (§4.2 is arguably the right citation for both, since it defines the option format itself.)
Why these are worth a follow-up rather than a shrug
#509 was verified behaviour-neutral by byte-identity (md5 d797c44b0c1b on both trees), identical test counts (466 passed / 1260 subtests), and 240 independent arithmetic comparisons with 0 mismatches. Every one of those proofs tests that nothing changed — none of them tests whether the extraction was complete, which is exactly the gap items 1 and 2 sit in. Worth remembering for the next extraction refactor: a completeness check (grep the old pattern to zero, per file) is a different proof from a neutrality check.
Three loose ends from #509 (PR #509, merged as
27bb315d5). None is a behaviour regression — the refactor is byte-identical, verified — but each undercuts a claim the refactor itself makes. The CGA length defect it also exposed is filed separately as #512.1.
mh.py:2853still open-codes the arithmetic the helper was created to collectpcapkit/protocols/internet/mh.py:2730-2732states the intent:But
_read_opt_padat:2853still reads:where the equivalent site in
pcapkit/protocols/internet/hopopt.py:603was converted:ipv6_opts.py:612-615was converted too. Only themh.pybranch was left behind, so the arithmetic still exists in two places in the file that introduced the helper.2. A cross-reference that points at nothing
mh.py:2727-2730refers the reader to "the surviving explanation of that fix … below". ThatNote:exists only inhopopt.py:579-588andipv6_opts.py:590-599;mh.py's_read_opt_padhas noNote:at all. The paragraph was copied and its "below" did not travel with it.3.
ipv6_opts.py:467cites the wrong RFC sectionBoth helpers cite
:rfc:8200#section-4.3``:hopopt.py:456— correct, §4.3 is the Hop-by-Hop Options Headeripv6_opts.py:467— wrong, IPv6-Opts is Destination Options, §4.6The same file already uses
#section-4.6correctly for the same kind of claim at:1253. Introduced by #509's own new docstring, copied verbatim fromhopopt.py. (§4.2 is arguably the right citation for both, since it defines the option format itself.)Why these are worth a follow-up rather than a shrug
#509 was verified behaviour-neutral by byte-identity (
md5 d797c44b0c1bon both trees), identical test counts (466 passed / 1260 subtests), and 240 independent arithmetic comparisons with 0 mismatches. Every one of those proofs tests that nothing changed — none of them tests whether the extraction was complete, which is exactly the gap items 1 and 2 sit in. Worth remembering for the next extraction refactor: a completeness check (grep the old pattern to zero, per file) is a different proof from a neutrality check.