# The skips an E2E gate run (IVG_GATE=1, scripts/ci-parity.sh) accepts: 4.6.0 R309, tests/e2e/gate.py.
#
# Every other skip under tests/e2e fails the gate run. One line per accepted skip:
#
#   <node id fnmatch pattern> | <text the skip message contains> | <why this skip is acceptable in the gate>
#
# The pattern names a file or tests in it (it must contain ".py"); the message text is matched as a substring of
# the skip message, so a test listed for one reason still fails when it skips for another. Never list a skip that
# means the gate did not run: the container not running, SKIP_IRIS_TESTS, a scratch namespace not provided
# (tests/unit/test_460_r309_gate_skips.py refuses those). A new line is a review decision: say why in the third field.
#
# Reviewed 2026-11-06 (R309). The list starts short on purpose; the first gate run lists every other skip.

tests/e2e/test_embedded_wgproto_e2e.py::* | set RUN_EMBEDDED_WGPROTO_SUITE=true | opt-in suite: the cases need a wgproto (CSP/MCP) request context that docker-exec compiled methods do not have; run with RUN_EMBEDDED_WGPROTO_SUITE=true on an AI Hub build
tests/e2e/test_230_harness_identity.py::test_another_running_containers_identity_is_rejected | no second IRIS container | needs a second IRIS container to be running beside the test container; the gate runs one
tests/e2e/test_460_r286_nkg_kept_dictionary_e2e.py::* | an Arno library is configured here | the kept dictionary is made only where no Arno library is configured; the enterprise test container has one, so this half runs on Community only
tests/e2e/test_stress_ingest.py::TestBulkIngestEdges::test_bulk_ingest_throughput_10k | throughput bound not checked under the parallel gate | the sharded gate (IVG_GATE_PARALLEL=1) runs its shards' IRIS containers side by side on one host, so a wall-clock rate there says nothing about IVG; the 10,000 edges are still ingested and must succeed, and the speed gate (scripts/bench/schema_gate.py) measures performance on a quiet host
tests/e2e/test_fraud_ui.py::* | agent-browser CLI not found | it drives the demo app at 127.0.0.1:8200 with the agent-browser CLI, and no fixture starts that app (one fixed port could not serve parallel shards either): run it by hand against a running demo server
tests/e2e/test_biomedical_ui.py::* | agent-browser CLI not found | it drives the demo app at 127.0.0.1:8200 with the agent-browser CLI, and no fixture starts that app (one fixed port could not serve parallel shards either): run it by hand against a running demo server
tests/e2e/test_hla_kg_e2e.py::* | set RUN_HLA_SUITE=true to enable | needs the private Mindwalk data (examples/expanded_mindwalk_KG_10000.*, from iris-vector-graph-private), which stays off shared gate hosts: the lead runs this file on the laptop (ivg-iris-enterprise) on the final SHA
tests/e2e/test_lazy_node_resolution.py::TestLazyNodeResolution::test_sc001_xl_cold_start_under_500ms | XL data not loaded | a cold-start timing bound on a 1M-edge load: the module loads a 1,110-edge tree for its answer checks, and a timing measured under a full suite's load is no benchmark (TESTING_POLICY, Environment); run it by hand after a 1M-edge BulkIngestEdges load
tests/e2e/test_embedded_wgproto_e2e.py::TestEmbeddedCommunityIRIS::* | iris.sql not available in Community container | Community's embedded Python has no iris.sql, which these cases need to run an embedded query; the enterprise half (TestEmbeddedConnectionRealIRIS) runs them
