test_result: Tolerate API errors when marking a test running

Sometimes these API fail, but that's not so important, what matters is marking commits as success or failure at the end of test.

bt5/erp5_test_result/DocumentTemplateItem/portal_components/document.erp5.GitlabRESTConnector.py

@@ -102,6 +102,13 @@ class GitlabRESTConnector(XMLObject):
       # branch. This typically happen after a new commit was push-forced to
       # the tested branch.
       return
+    if state == 'running' and response.status_code == requests.codes.bad_request:
+      # Since we updated to gitlab 9.5.10, sometimes annotating a test as running
+      # fail with "Cannot transition status via :run from :running" error. This
+      # seem to happen when previous test did not receive the finish notification.
+      # In that case, it's better to ignore the error, the commit will not be
+      # annotated as "running", but that seems acceptable.
+      return
     response.raise_for_status()
 
 

bt5/erp5_test_result/TestTemplateItem/portal_components/test.erp5.testTaskDistribution.py

@@ -1720,3 +1720,13 @@ class TestGitlabRESTConnectorInterface(ERP5TypeTestCase):
       self.test_result.start()
       self.tic()
 
+  def test_start_test_tolerate_errors(self):
+    with responses.RequestsMock() as rsps:
+      rsps.add(
+          responses.POST,
+          self.post_commit_status_url,
+          json={"message": 'Cannot transition status via :run from :running (Reason(s): Status cannot transition via "run")'},
+          status=httplib.BAD_REQUEST,
+      )
+      self.test_result.start()
+      self.tic()