diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 75b669f..7d5c4cd 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -58,9 +58,9 @@ jobs:
     - run: pip freeze
 
     - name: test with extras
-      run: |
-        make test
-        coverage xml
+      run: make test
+
+    - run: coverage xml
 
     - uses: codecov/codecov-action@v2.0.3
       with:
@@ -71,9 +71,9 @@ jobs:
       run: pip uninstall -y multidict numpy pydantic asyncpg
 
     - name: test without extras
-      run: |
-        make test
-        coverage xml
+      run: make test
+
+    - run: coverage xml
 
     - uses: codecov/codecov-action@v2.0.3
       with:
diff --git a/devtools/debug.py b/devtools/debug.py
index b9f7182..1954bf5 100644
--- a/devtools/debug.py
+++ b/devtools/debug.py
@@ -141,15 +141,16 @@ def _process(self, args, kwargs) -> DebugOutput:
                 warning=self._show_warnings and 'error parsing code, call stack too shallow',
             )
 
-        filename = call_frame.f_code.co_filename
         function = call_frame.f_code.co_name
-        if filename.startswith('/'):
-            # make the path relative
-            from pathlib import Path
 
+        from pathlib import Path
+
+        path = Path(call_frame.f_code.co_filename)
+        if path.is_absolute():
+            # make the path relative
             cwd = Path('.').resolve()
             try:
-                filename = str(Path(filename).relative_to(cwd))
+                path = path.relative_to(cwd)
             except ValueError:
                 # happens if filename path is not within CWD
                 pass
@@ -173,7 +174,7 @@ def _process(self, args, kwargs) -> DebugOutput:
                 arguments = list(self._process_args(ex, args, kwargs))
 
         return self.output_class(
-            filename=filename,
+            filename=str(path),
             lineno=lineno,
             frame=function,
             arguments=arguments,
diff --git a/tests/test_expr_render.py b/tests/test_expr_render.py
index 7d66268..9ff5fa6 100644
--- a/tests/test_expr_render.py
+++ b/tests/test_expr_render.py
@@ -1,21 +1,21 @@
 import asyncio
-import re
 import sys
 
 import pytest
 
 from devtools import Debug, debug
 
+from .utils import normalise_output
+
 
 def foobar(a, b, c):
     return a + b + c
 
 
-@pytest.mark.xfail(sys.platform == 'win32', reason='as yet unknown windows problem')
 def test_simple():
     a = [1, 2, 3]
     v = debug.format(len(a))
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     # print(s)
     assert (
         'tests/test_expr_render.py:<line no> test_simple\n'
@@ -23,18 +23,16 @@ def test_simple():
     ) == s
 
 
-@pytest.mark.xfail(sys.platform == 'win32', reason='as yet unknown windows problem')
 def test_subscription():
     a = {1: 2}
     v = debug.format(a[1])
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     assert (
         'tests/test_expr_render.py:<line no> test_subscription\n'
         '    a[1]: 2 (int)'
     ) == s
 
 
-@pytest.mark.xfail(sys.platform == 'win32', reason='as yet unknown windows problem')
 def test_exotic_types():
     aa = [1, 2, 3]
     v = debug.format(
@@ -49,8 +47,7 @@ def test_exotic_types():
         {a: a + 1 for a in aa},
         (a for a in aa),
     )
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
-    s = re.sub(r'(at 0x)\w+', r'\1<hash>', s)
+    s = normalise_output(str(v))
     print('\n---\n{}\n---'.format(v))
 
     # Generator expression source changed in 3.8 to include parentheses, see:
@@ -83,11 +80,10 @@ def test_exotic_types():
     ) == s
 
 
-@pytest.mark.xfail(sys.platform == 'win32', reason='as yet unknown windows problem')
 def test_newline():
     v = debug.format(
         foobar(1, 2, 3))
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     # print(s)
     assert (
         'tests/test_expr_render.py:<line no> test_newline\n'
@@ -95,12 +91,11 @@ def test_newline():
     ) == s
 
 
-@pytest.mark.xfail(sys.platform == 'win32', reason='as yet unknown windows problem')
 def test_trailing_bracket():
     v = debug.format(
         foobar(1, 2, 3)
     )
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     # print(s)
     assert (
         'tests/test_expr_render.py:<line no> test_trailing_bracket\n'
@@ -108,14 +103,13 @@ def test_trailing_bracket():
     ) == s
 
 
-@pytest.mark.xfail(sys.platform == 'win32', reason='as yet unknown windows problem')
 def test_multiline():
     v = debug.format(
         foobar(1,
                2,
                3)
     )
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     # print(s)
     assert (
         'tests/test_expr_render.py:<line no> test_multiline\n'
@@ -123,12 +117,11 @@ def test_multiline():
     ) == s
 
 
-@pytest.mark.xfail(sys.platform == 'win32', reason='as yet unknown windows problem')
 def test_multiline_trailing_bracket():
     v = debug.format(
         foobar(1, 2, 3
                ))
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     # print(s)
     assert (
         'tests/test_expr_render.py:<line no> test_multiline_trailing_bracket\n'
@@ -136,7 +129,6 @@ def test_multiline_trailing_bracket():
     ) == s
 
 
-@pytest.mark.xfail(sys.platform == 'win32', reason='as yet unknown windows problem')
 @pytest.mark.skipif(sys.version_info < (3, 6), reason='kwarg order is not guaranteed for 3.5')
 def test_kwargs():
     v = debug.format(
@@ -144,7 +136,7 @@ def test_kwargs():
         a=6,
         b=7
     )
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     assert (
         'tests/test_expr_render.py:<line no> test_kwargs\n'
         '    foobar(1, 2, 3): 6 (int)\n'
@@ -153,7 +145,6 @@ def test_kwargs():
     ) == s
 
 
-@pytest.mark.xfail(sys.platform == 'win32', reason='as yet unknown windows problem')
 @pytest.mark.skipif(sys.version_info < (3, 6), reason='kwarg order is not guaranteed for 3.5')
 def test_kwargs_multiline():
     v = debug.format(
@@ -162,7 +153,7 @@ def test_kwargs_multiline():
         a=6,
         b=7
     )
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     assert (
         'tests/test_expr_render.py:<line no> test_kwargs_multiline\n'
         '    foobar(1, 2, 3): 6 (int)\n'
@@ -171,20 +162,18 @@ def test_kwargs_multiline():
     ) == s
 
 
-@pytest.mark.xfail(sys.platform == 'win32', reason='as yet unknown windows problem')
 def test_multiple_trailing_lines():
     v = debug.format(
         foobar(
             1, 2, 3
         ),
     )
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     assert (
         'tests/test_expr_render.py:<line no> test_multiple_trailing_lines\n    foobar( 1, 2, 3 ): 6 (int)'
     ) == s
 
 
-@pytest.mark.xfail(sys.platform == 'win32', reason='as yet unknown windows problem')
 def test_very_nested_last_statement():
     def func():
         return debug.format(
@@ -201,14 +190,13 @@ def func():
 
     v = func()
     # check only the original code is included in the warning
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     assert s == (
         'tests/test_expr_render.py:<line no> func\n'
         '    abs( abs( abs( abs( -1 ) ) ) ): 1 (int)'
     )
 
 
-@pytest.mark.xfail(sys.platform == 'win32', reason='as yet unknown windows problem')
 def test_syntax_warning():
     def func():
         return debug.format(
@@ -227,7 +215,7 @@ def func():
 
     v = func()
     # check only the original code is included in the warning
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     assert s == (
         'tests/test_expr_render.py:<line no> func\n'
         '    abs( abs( abs( abs( abs( -1 ) ) ) ) ): 1 (int)'
@@ -254,14 +242,13 @@ def func():
         )
 
     v = func()
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     assert s == (
         'tests/test_expr_render.py:<line no> func\n'
         '    abs( abs( abs( abs( abs( -1 ) ) ) ) ): 1 (int)'
     )
 
 
-@pytest.mark.xfail(sys.platform == 'win32', reason='as yet unknown windows problem')
 def test_await():
     async def foo():
         return 1
@@ -272,7 +259,7 @@ async def bar():
     loop = asyncio.new_event_loop()
     v = loop.run_until_complete(bar())
     loop.close()
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     assert (
         'tests/test_expr_render.py:<line no> bar\n'
         '    await foo(): 1 (int)'
@@ -284,7 +271,7 @@ def test_other_debug_arg():
     v = debug.format([1, 2])
 
     # check only the original code is included in the warning
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     assert s == (
         'tests/test_expr_render.py:<line no> test_other_debug_arg\n'
         '    [1, 2] (list) len=2'
@@ -297,7 +284,7 @@ def test_other_debug_arg_not_literal():
     y = 2
     v = debug.format([x, y])
 
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     assert s == (
         'tests/test_expr_render.py:<line no> test_other_debug_arg_not_literal\n'
         '    [x, y]: [1, 2] (list) len=2'
@@ -310,7 +297,7 @@ def test_executing_failure():
     y = 2
 
     # executing fails inside a pytest assert ast the AST is modified
-    assert re.sub(r':\d{2,}', ':<line no>', str(debug.format([x, y]))) == (
+    assert normalise_output(str(debug.format([x, y]))) == (
         'tests/test_expr_render.py:<line no> test_executing_failure '
         '(executing failed to find the calling node)\n'
         '    [1, 2] (list) len=2'
@@ -326,7 +313,7 @@ def test_format_inside_error():
     except RuntimeError as e:
         v = str(e)
 
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     assert s == (
         'tests/test_expr_render.py:<line no> test_format_inside_error\n'
         '    [x, y]: [1, 2] (list) len=2'
diff --git a/tests/test_main.py b/tests/test_main.py
index 9b51b72..667a79a 100644
--- a/tests/test_main.py
+++ b/tests/test_main.py
@@ -9,7 +9,7 @@
 from devtools import Debug, debug
 from devtools.ansi import strip_ansi
 
-pytestmark = pytest.mark.xfail(sys.platform == 'win32', reason='as yet unknown windows problem')
+from .utils import normalise_output
 
 
 def test_print(capsys):
@@ -18,7 +18,7 @@ def test_print(capsys):
     result = debug(a, b)
     stdout, stderr = capsys.readouterr()
     print(stdout)
-    assert re.sub(r':\d{2,}', ':<line no>', stdout) == (
+    assert normalise_output(stdout) == (
         'tests/test_main.py:<line no> test_print\n'
         '    a: 1 (int)\n'
         '    b: 2 (int)\n'
@@ -33,7 +33,7 @@ def test_print_kwargs(capsys):
     result = debug(a, b, foo=[1, 2, 3])
     stdout, stderr = capsys.readouterr()
     print(stdout)
-    assert re.sub(r':\d{2,}', ':<line no>', stdout) == (
+    assert normalise_output(stdout) == (
         'tests/test_main.py:<line no> test_print_kwargs\n'
         '    a: 1 (int)\n'
         '    b: 2 (int)\n'
@@ -49,7 +49,7 @@ def test_print_generator(capsys):
     result = debug(gen)
     stdout, stderr = capsys.readouterr()
     print(stdout)
-    assert re.sub(r':\d{2,}', ':<line no>', stdout) == (
+    assert normalise_output(stdout) == (
         'tests/test_main.py:<line no> test_print_generator\n'
         '    gen: (\n'
         '        1,\n'
@@ -66,7 +66,7 @@ def test_format():
     a = b'i might bite'
     b = "hello this is a test"
     v = debug.format(a, b)
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     print(s)
     assert s == (
         "tests/test_main.py:<line no> test_format\n"
@@ -75,6 +75,10 @@ def test_format():
     )
 
 
+@pytest.mark.xfail(
+    sys.platform == 'win32',
+    reason='Fatal Python error: _Py_HashRandomization_Init: failed to get random numbers to initialize Python',
+)
 def test_print_subprocess(tmpdir):
     f = tmpdir.join('test.py')
     f.write("""\
@@ -109,7 +113,12 @@ def test_odd_path(mocker):
     mocked_relative_to = mocker.patch('pathlib.Path.relative_to')
     mocked_relative_to.side_effect = ValueError()
     v = debug.format('test')
-    assert re.search(r"/.*?/test_main.py:\d{2,} test_odd_path\n    'test' \(str\) len=4", str(v)), v
+    if sys.platform == "win32":
+        pattern = r"\w:\\.*?\\"
+    else:
+        pattern = r"/.*?/"
+    pattern += r"test_main.py:\d{2,} test_odd_path\n    'test' \(str\) len=4"
+    assert re.search(pattern, str(v)), v
 
 
 def test_small_call_frame():
@@ -119,7 +128,7 @@ def test_small_call_frame():
         2,
         3,
     )
-    assert re.sub(r':\d{2,}', ':<line no>', str(v)) == (
+    assert normalise_output(str(v)) == (
         'tests/test_main.py:<line no> test_small_call_frame\n'
         '    1 (int)\n'
         '    2 (int)\n'
@@ -135,7 +144,7 @@ def test_small_call_frame_warning():
         3,
     )
     print('\n---\n{}\n---'.format(v))
-    assert re.sub(r':\d{2,}', ':<line no>', str(v)) == (
+    assert normalise_output(str(v)) == (
         'tests/test_main.py:<line no> test_small_call_frame_warning\n'
         '    1 (int)\n'
         '    2 (int)\n'
@@ -147,7 +156,7 @@ def test_small_call_frame_warning():
 def test_kwargs():
     a = 'variable'
     v = debug.format(first=a, second='literal')
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     print(s)
     assert s == (
         "tests/test_main.py:<line no> test_kwargs\n"
@@ -160,7 +169,7 @@ def test_kwargs_orderless():
     # for python3.5
     a = 'variable'
     v = debug.format(first=a, second='literal')
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     assert set(s.split('\n')) == {
         "tests/test_main.py:<line no> test_kwargs_orderless",
         "    first: 'variable' (str) len=8 variable=a",
@@ -170,14 +179,14 @@ def test_kwargs_orderless():
 
 def test_simple_vars():
     v = debug.format('test', 1, 2)
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     assert s == (
         "tests/test_main.py:<line no> test_simple_vars\n"
         "    'test' (str) len=4\n"
         "    1 (int)\n"
         "    2 (int)"
     )
-    r = re.sub(r':\d{2,}', ':<line no>', repr(v))
+    r = normalise_output(repr(v))
     assert r == (
         "<DebugOutput tests/test_main.py:<line no> test_simple_vars arguments: 'test' (str) len=4 1 (int) 2 (int)>"
     )
@@ -239,17 +248,17 @@ def test_exec(capsys):
 
 def test_colours():
     v = debug.format(range(6))
-    s = re.sub(r':\d{2,}', ':<line no>', v.str(True))
+    s = v.str(True)
     assert s.startswith('\x1b[35mtests'), repr(s)
-    s2 = strip_ansi(s)
-    assert s2 == v.str(), repr(s2)
+    s2 = normalise_output(strip_ansi(s))
+    assert s2 == normalise_output(v.str()), repr(s2)
 
 
 def test_colours_warnings(mocker):
     mocked_getframe = mocker.patch('sys._getframe')
     mocked_getframe.side_effect = ValueError()
     v = debug.format('x')
-    s = re.sub(r':\d{2,}', ':<line no>', v.str(True))
+    s = normalise_output(v.str(True))
     assert s.startswith('\x1b[35m<unknown>'), repr(s)
     s2 = strip_ansi(s)
     assert s2 == v.str(), repr(s2)
@@ -270,10 +279,14 @@ def test_breakpoint(mocker):
     assert mocked_set_trace.called
 
 
+@pytest.mark.xfail(
+    sys.platform == 'win32' and sys.version_info >= (3, 9),
+    reason='see https://github.com/alexmojaki/executing/issues/27',
+)
 def test_starred_kwargs():
     v = {'foo': 1, 'bar': 2}
     v = debug.format(**v)
-    s = re.sub(r':\d{2,}', ':<line no>', v.str())
+    s = normalise_output(v.str())
     assert set(s.split('\n')) == {
         'tests/test_main.py:<line no> test_starred_kwargs',
         '    foo: 1 (int)',
@@ -289,11 +302,10 @@ def __getattr__(self, item):
 
     b = BadPretty()
     v = debug.format(b)
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
-    s = re.sub(r'0x[0-9a-f]+', '0x000', s)
+    s = normalise_output(str(v))
     assert s == (
         "tests/test_main.py:<line no> test_pretty_error\n"
-        "    b: <tests.test_main.test_pretty_error.<locals>.BadPretty object at 0x000> (BadPretty)\n"
+        "    b: <tests.test_main.test_pretty_error.<locals>.BadPretty object at 0x<hash>> (BadPretty)\n"
         "    !!! error pretty printing value: RuntimeError('this is an error')"
     )
 
@@ -302,7 +314,7 @@ def test_multiple_debugs():
     debug.format([i * 2 for i in range(2)])
     debug.format([i * 2 for i in range(2)])
     v = debug.format([i * 2 for i in range(2)])
-    s = re.sub(r':\d{2,}', ':<line no>', str(v))
+    s = normalise_output(str(v))
     assert s == (
         'tests/test_main.py:<line no> test_multiple_debugs\n'
         '    [i * 2 for i in range(2)]: [0, 2] (list) len=2'
diff --git a/tests/test_prettier.py b/tests/test_prettier.py
index 88b5ec3..cdabcd6 100644
--- a/tests/test_prettier.py
+++ b/tests/test_prettier.py
@@ -354,7 +354,8 @@ def test_cimultidict():
 def test_os_environ():
     v = pformat(os.environ)
     assert v.startswith('<_Environ({')
-    assert "    'HOME': '" in v
+    for key in os.environ:
+        assert f"    '{key}': " in v
 
 
 class Foo:
diff --git a/tests/test_utils.py b/tests/test_utils.py
index 9c455b0..d7c5681 100644
--- a/tests/test_utils.py
+++ b/tests/test_utils.py
@@ -43,8 +43,4 @@ def test_use_highlight_auto_win(monkeypatch):
     monkeypatch.delenv('TEST_DONT_USE_HIGHLIGHT', raising=False)
     monkeypatch.setattr(devtools.utils, 'isatty', lambda _=None: True)
 
-    monkeypatch.setattr(devtools.utils, 'color_active', False)
-    assert use_highlight() is False
-
-    monkeypatch.setattr(devtools.utils, 'color_active', True)
     assert use_highlight() is True
diff --git a/tests/utils.py b/tests/utils.py
new file mode 100644
index 0000000..d254c63
--- /dev/null
+++ b/tests/utils.py
@@ -0,0 +1,8 @@
+import re
+
+
+def normalise_output(s):
+    s = re.sub(r':\d{2,}', ':<line no>', s)
+    s = re.sub(r'(at 0x)\w+', r'\1<hash>', s)
+    s = s.replace('\\', '/')
+    return s