From 0a7c491e55e0349c97590007ff32d76e25a6bd1b Mon Sep 17 00:00:00 2001 From: Lilleman auf Larv Date: Sun, 6 Sep 2026 18:03:55 +0200 Subject: [PATCH] Say what the capture check really requires, and what it cannot express yet --- interop-tests/AGENTS.md | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/interop-tests/AGENTS.md b/interop-tests/AGENTS.md index 8688efb..42ffdb8 100644 --- a/interop-tests/AGENTS.md +++ b/interop-tests/AGENTS.md @@ -31,7 +31,14 @@ docker compose -f compose.yaml -f interop-tests/compose..yaml down -v and then decodes `captures/.pcapng` with tshark (`-d tcp.port==,smpp -Y smpp -T json`), printing the command histogram and the counts of `_ws.malformed` and error-severity `_ws.expert`. -Both counts must be zero for a phase to pass. +Both counts must be zero for a phase to pass, and an empty capture, or one carrying no bind and its +response, fails too. + +A peer whose scenarios send malformed PDUs on purpose — jsmpp does, to prove they are refused — has +no way to say so, so its run exits non-zero every time and its findings file carries the count that +is expected. Give the runner an expected count per peer, and a deviation from it becomes the signal +that a bare threshold cannot be: today a third malformed frame appearing beside jsmpp's two +deliberate ones looks exactly like the two. ## Rules for an experiment