fix: align CI and task-status handling with final 1.39 semantics (#2098)

The CI pin 1.39.0-rc.0-b41225e was a stale mid-review build of core
#12252 that emitted an IN_PROGRESS task status which never shipped; pin
1.39.0-rc.1 instead. Parse the task status tolerantly — known values
map to InvertedIndexTaskStatus, unknown values pass through as plain
strings, since the spec declares the field an open-vocabulary string —
and pin that with a mock test. Upgrade the coupled-tokenization test to
the final join contract: an identical re-PUT while the task is in
flight returns 202 with the existing taskId and STARTED, verified live
against 1.39.0-rc.1 (post-completion re-PUT stays 200 NO_OP).

Author: g-despot

.github/workflows/main.yaml

@@ -29,7 +29,7 @@ env:
   WEAVIATE_135: 1.35.18
   WEAVIATE_136: 1.36.12
   WEAVIATE_137: 1.37.5-e0fe0d5.amd64
-  WEAVIATE_139: 1.39.0-rc.0-b41225e.amd64
+  WEAVIATE_139: 1.39.0-rc.1
 
 jobs:
   lint-and-format:

integration/test_collection_config.py

@@ -2824,6 +2824,16 @@ def test_property_reindex_coupled_tokenization_change(
     assert task.status == InvertedIndexTaskStatus.STARTED
     assert task.task_id is not None
 
+    # join contract: an identical re-PUT while the task is in flight returns the EXISTING task
+    join = collection.config.update_property_index(
+        "name", InvertedIndexType.SEARCHABLE, tokenization=Tokenization.FIELD
+    )
+    if join.status == InvertedIndexTaskStatus.STARTED:
+        assert join.task_id == task.task_id
+    else:
+        # the task already finalized; a re-PUT of matching configuration is a no-op
+        assert join.status == InvertedIndexTaskStatus.NO_OP
+
     prop = next(p for p in collection.config.get_property_indexes().properties if p.name == "name")
     searchable = next(i for i in prop.indexes if i.type == "searchable")
     filterable = next(i for i in prop.indexes if i.type == "filterable")
@@ -2837,7 +2847,7 @@ def test_property_reindex_coupled_tokenization_change(
         # the task already finalized before the first poll
         assert searchable.tokenization == Tokenization.FIELD
 
-    # poll the status endpoint (via the wait path of a NO_OP upsert) until the migration is done
+    # poll the status endpoint (joining the in-flight task via the wait path) until done
     status = collection.config.update_property_index(
         "name",
         InvertedIndexType.SEARCHABLE,

mock_tests/test_property_reindex.py

@@ -134,6 +134,25 @@ def test_update_property_index_wait_for_completion(
     weaviate_139_mock.check_assertions()
 
 
+def test_update_property_index_tolerant_task_status(
+    weaviate_139_mock: HTTPServer, client_139: weaviate.WeaviateClient
+) -> None:
+    """Unknown task statuses pass through as raw strings (the spec declares the field open-vocabulary)."""
+    weaviate_139_mock.expect_request(
+        f"{SCHEMA_PATH}/properties/name/index/searchable",
+        method="PUT",
+        json={"tokenization": "word"},
+    ).respond_with_json({"taskId": TASK_ID, "status": "SOMETHING_NEW"}, status=202)
+
+    task = client_139.collections.use(COLLECTION).config.update_property_index(
+        "name", InvertedIndexType.SEARCHABLE, tokenization=Tokenization.WORD
+    )
+    assert task.task_id == TASK_ID
+    assert task.status == "SOMETHING_NEW"
+    assert not isinstance(task.status, InvertedIndexTaskStatus)
+    weaviate_139_mock.check_assertions()
+
+
 def test_update_property_index_bare_str_tenant(
     weaviate_139_mock: HTTPServer, client_139: weaviate.WeaviateClient
 ) -> None:

weaviate/collections/classes/config.py

@@ -2307,8 +2307,14 @@ class InvertedIndexState(str, BaseEnum):
 
 @dataclass
 class _InvertedIndexTask(_ConfigBase):
+    """A submitted property index task.
+
+    Known `status` values parse to `InvertedIndexTaskStatus`; unknown server values pass
+    through as plain strings, since the spec declares the field open-vocabulary.
+    """
+
     task_id: Optional[str]
-    status: InvertedIndexTaskStatus
+    status: Union[InvertedIndexTaskStatus, str]
 
 
 InvertedIndexTask = _InvertedIndexTask

weaviate/collections/classes/config_methods.py

@@ -568,9 +568,15 @@ def _references_from_config(schema: Dict[str, Any]) -> List[_ReferenceProperty]:
 
 
 def _inverted_index_task_from_json(response: Dict[str, Any]) -> _InvertedIndexTask:
+    raw_status = response["status"]
+    try:
+        status: Union[InvertedIndexTaskStatus, str] = InvertedIndexTaskStatus(raw_status)
+    except ValueError:
+        # the spec declares the field open-vocabulary; pass unknown values through
+        status = raw_status
     return _InvertedIndexTask(
         task_id=response.get("taskId"),
-        status=InvertedIndexTaskStatus(response["status"]),
+        status=status,
     )