fix(ci): multiply timeouts on windows platform (#2137)

The aim is to solve irrelevant failures on windows CI.
It often fails when waiting for study upgrades, in particular.
The PR re-activates the use of more workers (logical cores
instead of CPUs).

ANT-2043

---------

Signed-off-by: Sylvain Leclerc <[email protected]>

.github/workflows/main.yml

@@ -53,7 +53,7 @@ jobs:
           pip install -r requirements-dev.txt
       - name: Test with pytest
         run: |
-          pytest --cov antarest --cov-report xml -n auto
+          pytest --cov antarest --cov-report xml -n logical
       - name: Archive code coverage results
         if: matrix.os == 'ubuntu-20.04'
         uses: actions/upload-artifact@v4

tests/integration/prepare_proxy.py

@@ -90,7 +90,7 @@ def copy_study_and_upgrade(self, ref_study_id: str, target_version: int) -> str:
         task_id = res.json()
         assert task_id
 
-        task = wait_task_completion(self.client, self.user_access_token, task_id, timeout=20)
+        task = wait_task_completion(self.client, self.user_access_token, task_id, base_timeout=20)
         assert task.status == TaskStatus.COMPLETED
         return study_id
 
@@ -173,7 +173,7 @@ def generate_snapshot(self, variant_id: str, denormalize: bool = False, from_scr
         task_id = res.json()
         assert task_id
 
-        task = wait_task_completion(self.client, self.user_access_token, task_id, timeout=20)
+        task = wait_task_completion(self.client, self.user_access_token, task_id, base_timeout=20)
         assert task.status == TaskStatus.COMPLETED
 
     def create_area(self, study_id: str, *, name: str, country: str = "FR") -> t.Dict[str, t.Any]:

tests/integration/raw_studies_blueprint/test_download_matrices.py

@@ -53,7 +53,7 @@ def copy_upgrade_study(self, ref_study_id, target_version=820):
         task_id = res.json()
         assert task_id
 
-        task = wait_task_completion(self.client, self.user_access_token, task_id, timeout=20)
+        task = wait_task_completion(self.client, self.user_access_token, task_id, base_timeout=20)
         assert task.status == TaskStatus.COMPLETED
         return study_820_id
 
@@ -91,7 +91,7 @@ def generate_snapshot(self, variant_id: str, denormalize=False, from_scratch=Tru
         task_id = res.json()
         assert task_id
 
-        task = wait_task_completion(self.client, self.user_access_token, task_id, timeout=20)
+        task = wait_task_completion(self.client, self.user_access_token, task_id, base_timeout=20)
         assert task.status == TaskStatus.COMPLETED
 
     def create_area(self, parent_id, *, name: str, country: str = "FR") -> str:

tests/integration/utils.py

@@ -11,6 +11,7 @@
 # This file is part of the Antares project.
 
 import contextlib
+import os
 import time
 from typing import Callable
 
@@ -29,24 +30,32 @@ def wait_for(predicate: Callable[[], bool], timeout: float = 10, sleep_time: flo
     raise TimeoutError(f"task is still in progress after {timeout} seconds")
 
 
+IS_WINDOWS = os.name == "nt"
+TIMEOUT_MULTIPLIER = 2 if IS_WINDOWS else 1
+
+
 def wait_task_completion(
     client: TestClient,
     access_token: str,
     task_id: str,
     *,
-    timeout: float = 10,
+    base_timeout: float = 10,
 ) -> TaskDTO:
-    end_time = time.time() + timeout
-    while time.time() < end_time:
-        time.sleep(0.1)
-        res = client.request(
-            "GET",
-            f"/v1/tasks/{task_id}",
-            headers={"Authorization": f"Bearer {access_token}"},
-            json={"wait_for_completion": True},
-        )
-        assert res.status_code == 200
+    """
+    base_timeout is multiplied by 2 on windows to cope with slow CI
+    """
+    timeout = TIMEOUT_MULTIPLIER * base_timeout
+    params = {"wait_for_completion": True, "timeout": timeout}
+    res = client.request(
+        "GET",
+        f"/v1/tasks/{task_id}",
+        headers={"Authorization": f"Bearer {access_token}"},
+        params=params,
+    )
+    if res.status_code == 200:
         task = TaskDTO(**res.json())
         if task.status not in {TaskStatus.PENDING, TaskStatus.RUNNING}:
             return task
-    raise TimeoutError(f"{timeout} seconds")
+    elif res.status_code == 408:
+        raise TimeoutError(f"{timeout} seconds")
+    raise ValueError(f"Unexpected status code {res.status_code}")