Repository navigation
docs(generators): two false claims in options.py — ILNP nonce widths, and the SKIP padding rationale #1337
Description
Two comments in examples/generators/options.py state things that are not true. Both were found while cross-reviewing #1310 and verified against the tree.
1. :554-556 — the ILNP nonce claim is wrong on main already.
# ``nonce=0`` gives a zero-octet nonce. Only widths 3, 5, 6 and 7 pack
# at all, because a callable-length NumberField never clears the flag
# that says "hand struct a bytes"; 0xFFFFFF is three octets.
Enum_Option.ILNP_Nonce: {'nonce': 0xFFFFFF},Measured: nonce=0 gives a one-octet nonce, not zero. And widths 1, 2, 4 and 6 build and round-trip — on main as well as on #1310 — so "only widths 3, 5, 6 and 7 pack at all" is false in both directions. The override value 0xFFFFFF is fine; only the reasoning above it is wrong.
2. :1598-1604 — the SKIP rationale stops being true when #1310 merges.
The padding options are dropped by every
_make_*_optionsin the tree
After #1310 the IPv6 makers emit parsed padding in place, so this no longer holds for hopopt-option and ipv6-opts-option when the input is parsed data. The SKIP entries themselves stay correct — list input still drops padding, so Pad1 and PadN still produce byte-identical option-less headers — but the stated reason has to narrow to that case.
Blocked on #1310, which is what makes (2) wrong. (1) is independent.
Activity
JarryShaw commentedon Oct 7, 2026
One more stale comment for this issue, found while reviewing #1341:
examples/generators/options.py:1159-1161in_httpv2_rebuildsaysflagsis deliberately not passed on, becauseHTTP.makewould silently ignore it. Once #1341 merges,makeOR-combinesflagswith the frame's own, so that reason stops being true.JarryShaw commentedon Oct 8, 2026
Unblocked: #1310 and #1341 both merged on 2026-10-07. The stale comments are still at
options.py:561,:1628and:1159-1168ona8cc02594. A worker is on it now; it will check each claim against the code before rewriting it.fix(generators): correct stale options.py comments and the code they …