Show verification service errors

api/src/learn_to_cloud/routes/htmx_routes.py

@@ -100,6 +100,10 @@
     "Verification could not be started. This attempt was not counted — "
     "please try again."
 )
+_DURABLE_TERMINAL_ERROR_MESSAGE = (
+    "Verification failed because the verification service hit an internal error. "
+    "This attempt was not counted. Please try again."
+)
 
 _ACTIVE_DURABLE_STATUSES = {"pending", "running", "continuedasnew"}
 _TERMINAL_DURABLE_STATUSES = {"completed", "failed", "terminated", "canceled"}
@@ -155,7 +159,7 @@ async def _delete_terminal_job(
     request: Request,
     token_data: VerificationStatusToken,
     status: str,
-) -> None:
+) -> bool:
     """Drop the verification_jobs row after Durable reaches terminal failure.
 
     Replaces the older ``mark_server_error`` / ``mark_cancelled`` writes:
@@ -179,6 +183,7 @@ async def _delete_terminal_job(
             "verification.poller.delete_skipped",
             extra={"job_id": str(job_id), "runtime_status": status},
         )
+    return deleted
 
 
 async def _render_step_toggle(
@@ -570,8 +575,24 @@ async def htmx_verification_job_status(
         )
 
     if status in _DURABLE_FAILURE_STATUSES:
-        await _delete_terminal_job(request, token_data, status)
-        return HTMLResponse(_reload_verification_html())
+        deleted = await _delete_terminal_job(request, token_data, status)
+        if not deleted:
+            return HTMLResponse(_reload_verification_html())
+
+        requirement = await get_requirement_by_slug(db, token_data.requirement_slug)
+        if requirement is None:
+            return HTMLResponse(_reload_verification_html())
+
+        return templates.TemplateResponse(
+            request,
+            "partials/requirement_card.html",
+            build_requirement_card_context(
+                requirement=requirement,
+                github_username=current_user.github_username,
+                server_error=True,
+                server_error_message=_DURABLE_TERMINAL_ERROR_MESSAGE,
+            ),
+        )
 
     if status in _TERMINAL_DURABLE_STATUSES:
         return HTMLResponse(_reload_verification_html())

api/tests/routes/test_htmx_routes.py

@@ -656,10 +656,12 @@ async def test_completed_status_returns_reload_trigger(self):
         assert isinstance(result, HTMLResponse)
         assert "location.reload()" in bytes(result.body).decode()
 
-    async def test_failed_status_deletes_active_job(self):
+    async def test_failed_status_deletes_active_job_and_renders_error(
+        self,
+        _patch_templates,
+    ):
         """Durable terminal failure deletes the row instead of marking
-        a server-error status. Frees the partial unique index for retry
-        without copying Durable's terminal state into Postgres."""
+        a server-error status, then shows a retryable service error."""
         request = _mock_request()
         mock_session = AsyncMock()
         request.app.state.session_maker.return_value.__aenter__.return_value = (
@@ -688,6 +690,10 @@ async def test_failed_status_deletes_active_job(self):
                 "learn_to_cloud.routes.htmx_routes.VerificationJobRepository",
                 autospec=True,
             ) as mock_repository_class,
+            patch(
+                "learn_to_cloud.routes.htmx_routes.get_requirement_by_slug",
+                return_value=MagicMock(),
+            ),
         ):
             mock_repository = mock_repository_class.return_value
             mock_repository.delete_active = AsyncMock(return_value=True)
@@ -702,6 +708,13 @@ async def test_failed_status_deletes_active_job(self):
         assert isinstance(result, HTMLResponse)
         mock_repository.delete_active.assert_awaited_once_with(job_id)
         mock_session.commit.assert_awaited_once()
+        _, _, context = _patch_templates.TemplateResponse.call_args.args
+        assert context["server_error"] is True
+        assert (
+            context["server_error_message"]
+            == "Verification failed because the verification service hit an internal "
+            "error. This attempt was not counted. Please try again."
+        )
 
     async def test_canceled_status_also_deletes_active_job(self):
         """``Canceled`` and ``Terminated`` are handled the same way as
@@ -734,6 +747,10 @@ async def test_canceled_status_also_deletes_active_job(self):
                 "learn_to_cloud.routes.htmx_routes.VerificationJobRepository",
                 autospec=True,
             ) as mock_repository_class,
+            patch(
+                "learn_to_cloud.routes.htmx_routes.get_requirement_by_slug",
+                return_value=MagicMock(),
+            ),
         ):
             mock_repository = mock_repository_class.return_value
             mock_repository.delete_active = AsyncMock(return_value=True)