Retry dropped connections in dashboard API requests

.github/labview/test-dashboard-api.py

@@ -0,0 +1,48 @@
+import ast
+import http.client
+import io
+import json
+import pathlib
+import sys
+import time
+import unittest
+import urllib.error
+import urllib.request
+from unittest.mock import patch
+
+
+ROOT = pathlib.Path(__file__).resolve().parents[2]
+SOURCE = ROOT / 'actions' / 'dashboard' / 'dashboard.py'
+TREE = ast.parse(SOURCE.read_text(encoding='utf-8'))
+FUNCTIONS = [
+    node for node in TREE.body
+    if isinstance(node, ast.FunctionDef) and node.name in {'_retry_delay', 'gh_get'}
+]
+DASHBOARD = {
+    'json': json,
+    'repo': 'owner/repo',
+    'sys': sys,
+    'time': time,
+    'token': 'test-token',
+    'urllib': urllib,
+    '_API': {'degraded': False},
+}
+exec(compile(ast.Module(body=FUNCTIONS, type_ignores=[]), str(SOURCE), 'exec'), DASHBOARD)
+
+
+class DashboardApiTests(unittest.TestCase):
+    def test_remote_disconnect_retries_and_returns_response(self):
+        response = io.StringIO('{"ok": true}')
+        with patch('urllib.request.urlopen',
+                   side_effect=[http.client.RemoteDisconnected('connection dropped'), response]) as request:
+            with patch('time.sleep') as sleep:
+                result = DASHBOARD['gh_get']('commits/sha', _tries=2)
+
+        self.assertEqual(result, {'ok': True})
+        self.assertEqual(request.call_count, 2)
+        sleep.assert_called_once_with(1)
+        self.assertFalse(DASHBOARD['_API']['degraded'])
+
+
+if __name__ == '__main__':
+    unittest.main()

.github/pages/documentation.html

@@ -740,7 +740,8 @@ <h2 id="pages">9. The dashboard &amp; Pages site <a class="anchor" href="#pages"
         keeps paging back through history — past the recent window it retains <em>only</em> project revisions —
         until it has collected the project's revisions or reaches a large safety cap on how many commits it will
         classify. Reading that history is resilient to a transient GitHub&nbsp;API failure: rate-limited
-        (<code>HTTP&nbsp;403</code>/<code>429</code>), server, and network errors are retried with backoff
+        (<code>HTTP&nbsp;403</code>/<code>429</code>), server, and network errors (including dropped connections)
+        are retried with backoff
         (honoring <code>Retry-After</code>/<code>X-RateLimit-Reset</code>), and if a call still fails so the
         revision list would be incomplete, the build <strong>refuses to publish</strong> and exits &mdash; the
         previous good dashboard stays live and rebuilds automatically once the API recovers, so a burst of

actions/dashboard/dashboard.py

@@ -62,7 +62,7 @@ def gh_get(path, _tries=4):
             if transient:
                 _API['degraded'] = True
             return None
-        except (urllib.error.URLError, TimeoutError) as e:
+        except (urllib.error.URLError, OSError) as e:
             if attempt < n_tries - 1:
                 time.sleep(2 ** attempt)
                 continue

actions/dashboard/documentation.html

@@ -740,7 +740,8 @@ <h2 id="pages">9. The dashboard &amp; Pages site <a class="anchor" href="#pages"
         keeps paging back through history — past the recent window it retains <em>only</em> project revisions —
         until it has collected the project's revisions or reaches a large safety cap on how many commits it will
         classify. Reading that history is resilient to a transient GitHub&nbsp;API failure: rate-limited
-        (<code>HTTP&nbsp;403</code>/<code>429</code>), server, and network errors are retried with backoff
+        (<code>HTTP&nbsp;403</code>/<code>429</code>), server, and network errors (including dropped connections)
+        are retried with backoff
         (honoring <code>Retry-After</code>/<code>X-RateLimit-Reset</code>), and if a call still fails so the
         revision list would be incomplete, the build <strong>refuses to publish</strong> and exits &mdash; the
         previous good dashboard stays live and rebuilds automatically once the API recovers, so a burst of