Repository navigation
fix(internet): a standalone IGMPv3 Report cut inside a record warns 'packet length < 0', repeatedly #2749
Description
Describe the bug
Parse a standalone IGMPv3 Report whose capture is cut inside a group record. It emits SchemaWarning: packet length < 0, between one and five times for the same cut. Under the short-read convention, a cut is recorded on the layer as a short read and stays quiet. #1140 and #1148 fixed the same warning for other list items.
Reproduction
The #2738 round-5 review hit this on 2ade9f4b3: IGMPv3(data) on a Report with its last record cut mid-way. main (dea098b15) behaves the same, so #2738 does not cause it. Frames parsed through IPv4 are not affected.
Expected behavior
The cut reads TRUNCATED with its short-read record and emits no warning. The rebuild stays byte-exact.
Additional context
This is blocked until #2738 merges: that PR rewrites how a Report's record list is read (ListField count and shortfall), and the fix belongs on top of it. The warning comes from Schema.unpack's negative-length check.
Activity
JarryShaw commentedon Nov 3, 2026
#2738 has merged, so the earlier hold has cleared. This now waits on the #2534 regrouping, which is running in the serial window and moves every protocol module. It is dispatched as soon as that lands.
JarryShaw commentedon Nov 4, 2026
#2534 has landed (#2753), so this is unblocked. A fix is in progress on a branch off
1b4f9703e, in the shared Schema/field path rather than an IGMPv3-only suppression. It includes a sweep of other count-sized lists.JarryShaw commentedon Nov 4, 2026
The PR is #2755.
fix(corekit): read a counted list's cut item as a short read, not a n…