Categorize tests better

flake.nix

@@ -110,19 +110,29 @@
             status=$?
             if [ $status -eq 124 ]; then
                 result=timeout
-            elif [ $status -eq 101 ]; then
+            elif [ $status -eq 101 ] && ! grep -q '^// known-bug:' "$FILE"; then
                 result=panic
             elif [ $status -eq 0 ]; then
                 result=success
-                # Only keep the informative outputs.
-                rm "$FILE.charon-output"
             elif [ -f ${"$"}{FILE%.rs}.stderr ]; then
+                # This is a test that should fail
                 result=expected-failure
-                # Only keep the informative outputs.
-                rm "$FILE.charon-output"
+            elif grep -q '^// aux-build' "$FILE" \
+                  || grep -q '^// compile-flags:' "$FILE" \
+                  || grep -q '^// revisions:' "$FILE" \
+                  || grep -q '^// known-bug:' "$FILE" \
+                  || grep -q '^// edition:' "$FILE"; then
+                # We can't handle these for now
+                result=unsupported-build-settings
             else
                 result=failure
             fi
+
+            # Only keep the informative outputs.
+            if [[ $result != "panic" ]] && [[ $result != "failure" ]]; then
+                rm "$FILE.charon-output"
+            fi
+
             echo $result
           '';
           run_ui_tests = pkgs.writeScript "charon-analyze-test-file" ''