Ensure tempdir is kept for failed system tests

The temporary directory contains artifacts for the pytest module. That
module may contain multiple individual tests which were executed
sequentially. The artifacts should be kept if even one of these tests
failed.

Since pytest doesn't have any facility to expose test results to
fixtures, customize the pytest_runtest_makereport() hook to enable that.
It stores the test results into a session scope variable which is
available in all fixtures.

When deciding whether to remove the temporary directory, find the
relevant test results for this module and don't remove the tmpdir if any
one the tests failed.

bin/tests/system/conftest.py

@@ -142,6 +142,27 @@ if os.getenv("LEGACY_TEST_RUNNER", "0") == "0":
                 raise exc
             logging.debug(proc.stdout)
 
+    @pytest.hookimpl(tryfirst=True, hookwrapper=True)
+    def pytest_runtest_makereport(item):
+        """Hook that is used to expose test results to session (for use in fixtures)."""
+        # execute all other hooks to obtain the report object
+        outcome = yield
+        report = outcome.get_result()
+
+        # Set the test outcome in session, so we can access it from module-level
+        # fixture using nodeid. Note that this hook is called three times: for
+        # setup, call and teardown. We only care about the overall result so we
+        # merge the results together and preserve the information whether a test
+        # passed.
+        test_results = {}
+        try:
+            test_results = getattr(item.session, "test_results")
+        except AttributeError:
+            setattr(item.session, "test_results", test_results)
+        node_result = test_results.get(item.nodeid)
+        if node_result is None or report.outcome != "passed":
+            test_results[item.nodeid] = report.outcome
+
     # --------------------------- Fixtures -----------------------------------
 
     @pytest.fixture(scope="session")
@@ -240,6 +261,25 @@ if os.getenv("LEGACY_TEST_RUNNER", "0") == "0":
         FUTURE: This removes the need to have clean.sh scripts.
         """
 
+        def get_test_result():
+            """Aggregate test results from all individual tests from this module
+            into a single result: failed > skipped > passed."""
+            test_results = {
+                node.nodeid: request.session.test_results[node.nodeid]
+                for node in request.node.collect()
+                if node.nodeid in request.session.test_results
+            }
+            logger.debug(test_results)
+            assert len(test_results)
+            failed = any(res == "failed" for res in test_results.values())
+            skipped = any(res == "skipped" for res in test_results.values())
+            if failed:
+                return "failed"
+            if skipped:
+                return "skipped"
+            assert all(res == "passed" for res in test_results.values())
+            return "passed"
+
         # Create a temporary directory with a copy of the original system test dir contents
         system_test_root = Path(f"{env['TOP_BUILDDIR']}/bin/tests/system")
         testdir = Path(
@@ -266,9 +306,13 @@ if os.getenv("LEGACY_TEST_RUNNER", "0") == "0":
             os.chdir(old_cwd)
             logger.debug("changed workdir to: %s", old_cwd)
 
+            result = get_test_result()
+
             # Clean temporary dir unless it should be kept
             if request.config.getoption("--noclean"):
                 logger.debug("--noclean requested, keeping temporary directory")
+            elif result == "failed":
+                logger.debug("test failure detected, keeping temporary directory")
             else:
                 logger.debug("deleting temporary directory")
                 shutil.rmtree(testdir)