weed-cli 1.8.0__tar.gz → 1.8.2__tar.gz

{weed_cli-1.8.0/weed_cli.egg-info → weed_cli-1.8.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: weed-cli
-Version: 1.8.0
+Version: 1.8.2
 Summary: Censorship-resistant video PoC — discovery/hosting/download over signed relay events, a real Kademlia DHT, or a TLS-capable NAT-traversal tunnel
 License: MIT
 Keywords: p2p,video,censorship-resistant,discovery,dht,kademlia,nat-traversal

{weed_cli-1.8.0 → weed_cli-1.8.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "weed-cli"
-version = "1.8.0"
+version = "1.8.2"
 description = "Censorship-resistant video PoC — discovery/hosting/download over signed relay events, a real Kademlia DHT, or a TLS-capable NAT-traversal tunnel"
 readme = "README.md"
 license = { text = "MIT" }

weed_cli-1.8.2/tests/test_dht.py

@@ -0,0 +1,128 @@
+"""
+dht.py's real Kademlia swarm -- multiple real dht.DHTNode instances (each
+its own thread + its own asyncio event loop, see DHTNode's own docstring)
+in one process. Unlike discovery_relay.py, kademlia.network.Server carries
+no module-level global state shared across instances, so running several
+in-process (rather than as separate subprocesses, the way
+tests/conftest.py's `relay` fixture has to for discovery_relay.py) is
+correct here, not a test-isolation shortcut -- kademlia's own test suite
+does the same thing for the same reason.
+"""
+import pytest
+
+import dht
+from testutil import free_port
+
+
+def make_node(bootstrap_port=None):
+    port = free_port()
+    bootstrap = [('127.0.0.1', bootstrap_port)] if bootstrap_port else None
+    node = dht.DHTNode(port, bootstrap_nodes=bootstrap, quiet=True)
+    node.start()
+    return node, port
+
+
+@pytest.fixture()
+def swarm():
+    """Yields a list callers append (node, port) pairs to; stops every
+    node on teardown regardless of how many the test actually created."""
+    nodes = []
+    yield nodes
+    for node, _ in nodes:
+        node.stop()
+
+
+def test_solo_node_with_no_neighbors_cannot_store_or_retrieve_anything(swarm):
+    """Not a trivial "empty DHT" check -- a real, easy-to-miss operational
+    gotcha discovered while writing these tests: kademlia's get/set work
+    by routing to the nearest *known* nodes, and a node with zero known
+    neighbors (the very first node in a swarm, before anyone else has
+    bootstrapped to or from it) has nowhere to route to. It doesn't raise
+    -- it logs a warning and silently no-ops, so announce()/lookup() both
+    just look like "nothing here" instead of "not connected to anything
+    yet." A genuinely-empty lookup in a *working* multi-node swarm is
+    covered separately below (test_lookup_for_unknown_hash_on_a_real_swarm)."""
+    a, _ = make_node()
+    swarm.append((a, _))
+    a.announce('c' * 64, '127.0.0.1:9201')
+    assert a.lookup('c' * 64) == []
+
+
+def test_two_nodes_announce_and_lookup(swarm):
+    """The actual round trip: B joins A's swarm, B announces, A looks it
+    up -- A and B never talk except for B's one bootstrap connection, so
+    finding it on A proves the DHT's own get/set is doing the work, not
+    some direct A<->B shortcut."""
+    a, a_port = make_node()
+    swarm.append((a, a_port))
+    b, b_port = make_node(bootstrap_port=a_port)
+    swarm.append((b, b_port))
+
+    b.announce('c' * 64, '127.0.0.1:9201', title='My Video')
+
+    results = a.lookup('c' * 64)
+    assert results == [{'host': '127.0.0.1:9201', 'title': 'My Video'}]
+
+
+def test_lookup_for_unknown_hash_on_a_real_swarm_is_empty(swarm):
+    a, a_port = make_node()
+    swarm.append((a, a_port))
+    b, b_port = make_node(bootstrap_port=a_port)
+    swarm.append((b, b_port))
+
+    b.announce('c' * 64, '127.0.0.1:9201')
+    assert a.lookup('d' * 64) == []
+
+
+def test_two_hosts_for_same_content_hash_merge_not_overwrite(swarm):
+    """_announce's own contract (see its docstring): a second host
+    announcing the same content_hash must show up *alongside* the first,
+    not replace it -- multiple peers hosting the same file is real,
+    useful redundancy, not a conflict to resolve."""
+    a, a_port = make_node()
+    swarm.append((a, a_port))
+    b, b_port = make_node(bootstrap_port=a_port)
+    swarm.append((b, b_port))
+
+    a.announce('c' * 64, '127.0.0.1:9201', title='v')
+    b.announce('c' * 64, '127.0.0.1:9202', title='v')
+
+    hosts = {e['host'] for e in a.lookup('c' * 64)}
+    assert hosts == {'127.0.0.1:9201', '127.0.0.1:9202'}
+
+
+def test_reannouncing_same_host_replaces_its_own_entry_not_duplicates(swarm):
+    """A completely solo node (zero known neighbors -- nobody's ever
+    bootstrapped to or from it) can't even get/set against itself:
+    kademlia's own set/get work by routing to the nearest known nodes, and
+    with none known there's nowhere to route to (it logs a warning and
+    silently no-ops, discovered the hard way -- an earlier version of
+    this test tried exactly that and got [] back instead of real data).
+    Two nodes is the minimum for get/set to actually do anything."""
+    a, a_port = make_node()
+    swarm.append((a, a_port))
+    b, b_port = make_node(bootstrap_port=a_port)
+    swarm.append((b, b_port))
+
+    a.announce('c' * 64, '127.0.0.1:9201', title='v1')
+    a.announce('c' * 64, '127.0.0.1:9201', title='v2')
+
+    results = b.lookup('c' * 64)
+    assert results == [{'host': '127.0.0.1:9201', 'title': 'v2'}]
+
+
+def test_three_node_chain_propagates_through_indirect_bootstrap(swarm):
+    """C bootstraps through B, never touches A directly -- if C can still
+    look up something A announced, that's Kademlia's own routing doing
+    real work, not just a two-node special case."""
+    a, a_port = make_node()
+    swarm.append((a, a_port))
+    b, b_port = make_node(bootstrap_port=a_port)
+    swarm.append((b, b_port))
+    c, c_port = make_node(bootstrap_port=b_port)
+    swarm.append((c, c_port))
+
+    a.announce('c' * 64, '127.0.0.1:9201', title='from A')
+
+    results = c.lookup('c' * 64)
+    assert results == [{'host': '127.0.0.1:9201', 'title': 'from A'}]

{weed_cli-1.8.0 → weed_cli-1.8.2/weed_cli.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: weed-cli
-Version: 1.8.0
+Version: 1.8.2
 Summary: Censorship-resistant video PoC — discovery/hosting/download over signed relay events, a real Kademlia DHT, or a TLS-capable NAT-traversal tunnel
 License: MIT
 Keywords: p2p,video,censorship-resistant,discovery,dht,kademlia,nat-traversal

{weed_cli-1.8.0 → weed_cli-1.8.2}/weed_cli.egg-info/SOURCES.txt

@@ -10,6 +10,7 @@ shell.py
 tunnel_relay.py
 web_ui.py
 weed.py
+tests/test_dht.py
 tests/test_discovery_relay.py
 tests/test_node_manifest.py
 tests/test_web_ui_api.py