Fix get_artifact returning the TC API response instead of the artifact

This fixes a regression from 384afee
where we started using the taskcluster python client instead of making
requests manually.

The artifacts route is special because the API returns some JSON as part
of a 303 to the artifact's content URL. The previous code followed that
303. The python client does not. So instead we're getting the body which
looks like this: `{"type": "s3/...", "url": "whereverthe303pointsat"}`.

This fixed it by following the redirect manually to restore the previous
behavior.
I also fixed the test as the previous one was basically ignoring the
fact that taskcluster was returning a 303 and was mocking as if it
didn't exist.

It's worth noting that while it restores the behavior of returning the
artifact's content (as a ByteIO), it is still a breaking change as the
previous code used to return a Response object.

Fixes #812

Author: Eijebong

src/taskgraph/util/taskcluster.py

@@ -170,7 +170,8 @@ def get_artifact(task_id, path):
     """
     queue = get_taskcluster_client("queue")
     response = queue.getLatestArtifact(task_id, path)
-    return _handle_artifact(path, response)
+    artifact_response = requests.get(response["url"])
+    return _handle_artifact(path, artifact_response)
 
 
 def list_artifacts(task_id):

test/test_util_taskcluster.py

@@ -110,26 +110,44 @@ def test_get_artifact(responses, root_url):
 
     # Test text artifact
     responses.get(
-        f"{root_url}/api/queue/v1/task/{tid}/artifacts/artifact.txt",
+        f"http://foo.bar/artifact.txt",
         body=b"foobar",
     )
+    responses.get(
+        f"{root_url}/api/queue/v1/task/{tid}/artifacts/artifact.txt",
+        body=b'{"type": "s3", "url": "http://foo.bar/artifact.txt"}',
+        status=303,
+        headers={"Location": "http://foo.bar/artifact.txt"}
+    )
     raw = tc.get_artifact(tid, "artifact.txt")
     assert raw.read() == b"foobar"
 
     # Test JSON artifact
+    responses.get(
+        f"http://foo.bar/artifact.json",
+        json={"foo": "bar"}
+    )
     responses.get(
         f"{root_url}/api/queue/v1/task/{tid}/artifacts/artifact.json",
-        json={"foo": "bar"},
+        body=b'{"type": "s3", "url": "http://foo.bar/artifact.json"}',
+        status=303,
+        headers={"Location": "http://foo.bar/artifact.json"}
     )
     result = tc.get_artifact(tid, "artifact.json")
     assert result == {"foo": "bar"}
 
     # Test YAML artifact
     expected_result = {"foo": b"\xe2\x81\x83".decode()}
     responses.get(
-        f"{root_url}/api/queue/v1/task/{tid}/artifacts/artifact.yml",
+        f"http://foo.bar/artifact.yml",
         body=b'foo: "\xe2\x81\x83"',
     )
+    responses.get(
+        f"{root_url}/api/queue/v1/task/{tid}/artifacts/artifact.yml",
+        body=b'{"type": "s3", "url": "http://foo.bar/artifact.yml"}',
+        status=303,
+        headers={"Location": "http://foo.bar/artifact.yml"}
+    )
     result = tc.get_artifact(tid, "artifact.yml")
     assert result == expected_result