Prevent from overwriting global configurations (#105)

This fixes a test failure that started to occur after #101.

```
! _pytest.outcomes.Exit: playwright failed to launch
```

The cause of the problem was that the combination of the `pytester` fixture and the `pytest_configure` hook is not very good.
We use `pytest_configure` to manage global settings, but this function is called again when we run sub-tests using pytester, resulting in new global settings being overwritten.

So I added a check to prevent overwriting global settings once it is set.

pytest_pyodide/hook.py

@@ -92,9 +92,15 @@ def pytest_configure(config):
     )
 
     run_host, runtimes = _filter_runtimes(config.option.runtime)
-    pytest.pyodide_run_host_test = run_host
-    pytest.pyodide_runtimes = runtimes
-    pytest.pyodide_dist_dir = config.getoption("--dist-dir")
+
+    # using `pytester` fixture seems to call this hook again with different options
+    # so this is a workaround to avoid overwriting the values
+    if not hasattr(pytest, "pyodide_run_host_test"):
+        pytest.pyodide_run_host_test = run_host
+    if not hasattr(pytest, "pyodide_runtimes"):
+        pytest.pyodide_runtimes = runtimes
+    if not hasattr(pytest, "pyodide_dist_dir"):
+        pytest.pyodide_dist_dir = config.getoption("--dist-dir")
 
 
 @pytest.hookimpl(tryfirst=True)

tests/test_options.py

@@ -1,5 +1,7 @@
 import pytest
 
+from pytest_pyodide.hook import _filter_runtimes
+
 
 def test_dist_dir(pytester):
     dist_dir = "dist"
@@ -47,49 +49,40 @@ def test_option(request):
     assert result.ret == 4
 
 
-@pytest.mark.parametrize(
-    "_runtime",
-    [
-        "chrome",
-        "firefox",
-        "safari",
-        "node",
-        "firefox,chrome",
-    ],
-)
-def test_runtime(pytester, _runtime):
-    runtimes = _runtime.split(",")
-
-    pytester.makepyfile(
-        f"""
-        import pytest
-        def test_option():
-            assert pytest.pyodide_runtimes == set({runtimes!r})
-        """
-    )
-
-    result = pytester.runpytest("--runtime", _runtime)
-    result.assert_outcomes(passed=1)
-
-
 @pytest.mark.parametrize(
     "_runtime",
     [
         "invalid",
     ],
 )
 def test_invalid_runtime(pytester, _runtime):
-    runtimes = _runtime.split(",")
+    _runtime.split(",")
 
     pytester.makepyfile(
-        f"""
+        """
         import pytest
         def test_option():
-            assert pytest.pyodide_runtimes == set({runtimes!r})
+            assert True
         """
     )
 
     # TODO: catch internal errors directly?
     with pytest.raises(ValueError, match="Pytest terminal summary report not found"):
         result = pytester.runpytest("--runtime", _runtime)
         result.assert_outcomes(errors=1)
+
+
+@pytest.mark.parametrize(
+    "_runtime,expected",
+    [
+        ("chrome", (True, {"chrome"})),
+        ("firefox", (True, {"firefox"})),
+        ("node", (True, {"node"})),
+        ("chrome,firefox", (True, {"chrome", "firefox"})),
+        ("chrome-no-host,firefox", (False, {"chrome", "firefox"})),
+        ("host", (True, set())),
+        ("chrome-no-host,host", (True, {"chrome"})),
+    ],
+)
+def test_filter_runtimes(_runtime, expected):
+    _filter_runtimes(_runtime) == expected