diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 746dfa3aed..0c0283dec2 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -96,7 +96,6 @@ jobs:
       run: |
         tox -e py-core
         cd core
-        python -m coverage xml
         mv .coverage .coverage.${{ matrix.platform }}.${{ matrix.python-version }}
     - name: Store coverage data
       uses: actions/upload-artifact@v3
@@ -104,13 +103,6 @@ jobs:
         name: core-coverage-data
         path: "core/.coverage.*"
         if-no-files-found: error
-    - name: Upload coverage data to CodeCov
-      uses: codecov/codecov-action@v3
-      with:
-        token: ${{ secrets.CODECOV_TOKEN }}
-        file: core/coverage.xml
-        flags: unittests
-        fail_ci_if_error: true
 
   core-coverage:
     name: Combine & check core coverage.
@@ -190,3 +182,58 @@ jobs:
         TOGA_INSTALL_COMMAND: 'bash -c "pip install ../{core,dummy,${{ matrix.backend }}}/dist/*.whl"'
       run: |
         tox -e py-${{ matrix.backend }}
+
+  testbed:
+    needs: core
+    strategy:
+      fail-fast: false
+      matrix:
+        backend: ["macOS", "windows", "linux", "android", "iOS"]
+        include:
+          - pre-command:
+            briefcase-run-prefix:
+            briefcase-run-args:
+
+          - backend: macOS
+            runs-on: macos-12
+
+          - backend: linux
+            runs-on: ubuntu-latest
+            pre-command: "sudo apt-get update -y && sudo apt-get install -y python3-gi python3-gi-cairo gir1.2-gtk-3.0 python3-dev libgirepository1.0-dev libcairo2-dev pkg-config"
+            briefcase-run-prefix: 'xvfb-run -a -s "-screen 0 2048x1536x24"'
+
+          - backend: windows
+            runs-on: windows-latest
+
+          - backend: iOS
+            runs-on: macos-12
+            briefcase-run-args: ' -d "iPhone SE (3rd generation)"'
+
+          - backend: android
+            runs-on: macos-12
+            briefcase-run-args: " -d '{\"avd\":\"beePhone\"}' --Xemulator=-no-window --Xemulator=-no-snapshot --Xemulator=-no-audio --Xemulator=-no-boot-anim --shutdown-on-exit"
+
+    runs-on: ${{ matrix.runs-on }}
+    steps:
+    - uses: actions/checkout@v3.1.0
+      with:
+        fetch-depth: 0
+    - name: Set up Python
+      uses: actions/setup-python@v4.3.0
+      with:
+        python-version: "3.X"
+    - name: Install dependencies
+      run: |
+        ${{ matrix.pre-command }}
+        # Use the development version of Briefcase
+        pip install git+https://github.com/beeware/briefcase.git
+    - name: Test App
+      run: |
+        cd testbed
+        ${{ matrix.briefcase-run-prefix }} briefcase run ${{ matrix.backend }} --test ${{ matrix.briefcase-run-args }}
+    - uses: actions/upload-artifact@v3
+      name: Upload logs
+      if: failure()
+      with:
+        name: testbed-failure-logs-${{ matrix.backend }}
+        path: testbed/logs/*
diff --git a/android/MANIFEST.in b/android/MANIFEST.in
index 126da665ed..96861fad80 100644
--- a/android/MANIFEST.in
+++ b/android/MANIFEST.in
@@ -3,3 +3,4 @@ include LICENSE
 include README.rst
 include tox.ini
 recursive-include tests *.py
+recursive-include tests_backend *.py
diff --git a/android/tests_backend/__init__.py b/android/tests_backend/__init__.py
new file mode 100644
index 0000000000..e69de29bb2
diff --git a/android/tests_backend/widgets/__init__.py b/android/tests_backend/widgets/__init__.py
new file mode 100644
index 0000000000..e69de29bb2
diff --git a/android/tests_backend/widgets/base.py b/android/tests_backend/widgets/base.py
new file mode 100644
index 0000000000..bfd934d472
--- /dev/null
+++ b/android/tests_backend/widgets/base.py
@@ -0,0 +1,43 @@
+from pytest import skip
+
+
+class SimpleProbe:
+    def __init__(self, widget):
+        self.native = widget._impl.native
+        assert isinstance(self.native, self.native_class)
+
+    def assert_container(self, container):
+        container_native = container._impl.native
+        for i in range(container_native.getChildCount()):
+            child = container_native.getChildAt(i)
+            if child is self.native:
+                break
+        else:
+            raise AssertionError(f"cannot find {self.native} in {container_native}")
+
+    @property
+    def enabled(self):
+        return self.native.isEnabled()
+
+    @property
+    def background_color(self):
+        skip("not implemented: background_color")
+
+    @property
+    def color(self):
+        skip("not implemented: color")
+
+    @property
+    def hidden(self):
+        skip("not implemented: hidden")
+
+    @property
+    def width(self):
+        return self.native.getWidth()
+
+    @property
+    def height(self):
+        return self.native.getHeight()
+
+    def press(self):
+        self.native.performClick()
diff --git a/android/tests_backend/widgets/button.py b/android/tests_backend/widgets/button.py
new file mode 100644
index 0000000000..d707f8374d
--- /dev/null
+++ b/android/tests_backend/widgets/button.py
@@ -0,0 +1,8 @@
+from java import jclass
+
+from .label import LabelProbe
+
+
+# On Android, a Button is just a TextView with a state-dependent background image.
+class ButtonProbe(LabelProbe):
+    native_class = jclass("android.widget.Button")
diff --git a/android/tests_backend/widgets/label.py b/android/tests_backend/widgets/label.py
new file mode 100644
index 0000000000..553f52aa59
--- /dev/null
+++ b/android/tests_backend/widgets/label.py
@@ -0,0 +1,16 @@
+from java import jclass
+
+from .base import SimpleProbe
+from .properties import toga_color
+
+
+class LabelProbe(SimpleProbe):
+    native_class = jclass("android.widget.TextView")
+
+    @property
+    def color(self):
+        return toga_color(self.native.getCurrentTextColor())
+
+    @property
+    def text(self):
+        return str(self.native.getText())
diff --git a/android/tests_backend/widgets/properties.py b/android/tests_backend/widgets/properties.py
new file mode 100644
index 0000000000..95a56903dd
--- /dev/null
+++ b/android/tests_backend/widgets/properties.py
@@ -0,0 +1,15 @@
+from java import jint
+
+from android.graphics import Color
+from toga.colors import rgba
+
+
+def toga_color(color_int):
+    # Select the `int` overloads rather than the `long` ones.
+    color_int = jint(color_int)
+    return rgba(
+        Color.red(color_int),
+        Color.green(color_int),
+        Color.blue(color_int),
+        Color.alpha(color_int) / 255,
+    )
diff --git a/android/tests_backend/widgets/slider.py b/android/tests_backend/widgets/slider.py
new file mode 100644
index 0000000000..fefa45c154
--- /dev/null
+++ b/android/tests_backend/widgets/slider.py
@@ -0,0 +1,24 @@
+from java import jclass
+
+from android.os import Build
+
+from .base import SimpleProbe
+
+
+class SliderProbe(SimpleProbe):
+    native_class = jclass("android.widget.SeekBar")
+
+    @property
+    def position(self):
+        return (self.native.getProgress() - self._min) / (self._max - self._min)
+
+    def change(self, position):
+        self.native.setProgress(self._min + round(position * (self._max - self._min)))
+
+    @property
+    def _min(self):
+        return 0 if (Build.VERSION.SDK_INT < 26) else self.native.getMin()
+
+    @property
+    def _max(self):
+        return self.native.getMax()
diff --git a/changes/1687.misc.rst b/changes/1687.misc.rst
new file mode 100644
index 0000000000..66c7d93862
--- /dev/null
+++ b/changes/1687.misc.rst
@@ -0,0 +1 @@
+Added a backend test app.
diff --git a/iOS/src/toga_iOS/app.py b/iOS/src/toga_iOS/app.py
index 7dc5204253..7a2d5d8764 100644
--- a/iOS/src/toga_iOS/app.py
+++ b/iOS/src/toga_iOS/app.py
@@ -110,8 +110,10 @@ def open_document(self, fileURL):
         pass
 
     def main_loop(self):
-        # Main loop is a no-op on iOS; the app loop is integrated with the
-        # main iOS event loop.
+        # Main loop is non-blocking on iOS. The app loop is integrated with the
+        # main iOS event loop, so this call will return; however, it will leave
+        # the app in a state such that asyncio events will be scheduled on the
+        # iOS event loop.
         self.loop.run_forever_cooperatively(lifecycle=iOSLifecycle())
 
     def set_main_window(self, window):
diff --git a/pyproject.toml b/pyproject.toml
index 90f22ecd40..9ffe9d28a9 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -35,6 +35,7 @@ known_first_party = [
     "toga_dummy",
     "toga_gtk",
     "toga_iOS",
+    "toga_test",
     "toga_web",
     "toga_winforms",
 ]
diff --git a/testbed/.gitignore b/testbed/.gitignore
new file mode 100644
index 0000000000..34661cf5b9
--- /dev/null
+++ b/testbed/.gitignore
@@ -0,0 +1,69 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# OSX useful to ignore
+*.DS_Store
+.AppleDouble
+.LSOverride
+
+# Thumbnails
+._*
+
+# Files that might appear in the root of a volume
+.DocumentRevisions-V100
+.fseventsd
+.Spotlight-V100
+.TemporaryItems
+.Trashes
+.VolumeIcon.icns
+.com.apple.timemachine.donotpresent
+
+# Directories potentially created on remote AFP share
+.AppleDB
+.AppleDesktop
+Network Trash Folder
+Temporary Items
+.apdisk
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+env/
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+*.egg-info/
+.installed.cfg
+*.egg
+
+# IntelliJ Idea family of suites
+.idea
+*.iml
+## File-based project format:
+*.ipr
+*.iws
+## mpeltonen/sbt-idea plugin
+.idea_modules/
+
+# Briefcase build directories
+iOS/
+macOS/
+windows/
+android/
+linux/
+django/
+
+# Briefcase log files
+logs/
diff --git a/testbed/pyproject.toml b/testbed/pyproject.toml
new file mode 100644
index 0000000000..7bbe9ce98e
--- /dev/null
+++ b/testbed/pyproject.toml
@@ -0,0 +1,90 @@
+# This project was generated using template: https://github.com/beeware/briefcase-template and branch: v0.3.12
+[tool.briefcase]
+project_name = "Toga Testbed"
+bundle = "org.beeware.toga"
+version = "0.0.1"
+url = "https://beeware.org"
+license = "BSD license"
+author = 'Tiberius Yak'
+author_email = "tiberius@beeware.org"
+
+[tool.briefcase.app.testbed]
+formal_name = "Toga Testbed"
+description = "A testbed for Toga visual tests"
+icon = "src/testbed/resources/testbed"
+sources = [
+    'src/testbed',
+]
+test_sources = [
+    'tests',
+]
+requires = [
+    '../core',
+]
+test_requires = [
+    'pytest==7.2.0',
+    'pytest-asyncio==0.20.2',
+]
+
+
+[tool.briefcase.app.testbed.macOS]
+requires = [
+    '../cocoa',
+    'std-nslog~=1.0.0'
+]
+
+[tool.briefcase.app.testbed.linux]
+requires = [
+    '../gtk',
+]
+
+[tool.briefcase.app.testbed.linux.appimage]
+system_requires = [
+    'gir1.2-webkit-3.0',
+    'libcairo2-dev',
+    'libgirepository1.0-dev',
+    'libgtk-3-dev',
+    'libpango1.0-dev',
+    'librsvg2-dev',
+    'libwebkitgtk-3.0-0',
+]
+linuxdeploy_plugins = [
+    'DEPLOY_GTK_VERSION=3 gtk',
+]
+
+[tool.briefcase.app.testbed.linux.flatpak]
+flatpak_runtime = 'org.gnome.Platform'
+flatpak_runtime_version = '42'
+flatpak_sdk = 'org.gnome.Sdk'
+
+[tool.briefcase.app.testbed.windows]
+test_sources = [
+    '../winforms/tests_backend',
+]
+requires = [
+    '../winforms',
+]
+
+# Mobile deployments
+[tool.briefcase.app.testbed.iOS]
+requires = [
+    '../iOS',
+    'std-nslog~=1.0.0'
+]
+
+[tool.briefcase.app.testbed.android]
+test_sources = [
+    '../android/tests_backend',
+]
+requires = [
+    '../android'
+]
+
+# TODO: replace with extractPackages
+build_gradle_extra_content = "android.defaultConfig.python.pyc.src false"
+
+[tool.briefcase.app.testbed.web]
+requires = [
+    '../web'
+]
+style_framework = "Bootstrap v4.6"
diff --git a/testbed/src/testbed/__init__.py b/testbed/src/testbed/__init__.py
new file mode 100644
index 0000000000..e69de29bb2
diff --git a/testbed/src/testbed/__main__.py b/testbed/src/testbed/__main__.py
new file mode 100644
index 0000000000..7c4c4d4546
--- /dev/null
+++ b/testbed/src/testbed/__main__.py
@@ -0,0 +1,4 @@
+from testbed.app import main
+
+if __name__ == "__main__":
+    main().main_loop()
diff --git a/testbed/src/testbed/app.py b/testbed/src/testbed/app.py
new file mode 100644
index 0000000000..02f0f9499e
--- /dev/null
+++ b/testbed/src/testbed/app.py
@@ -0,0 +1,27 @@
+import toga
+
+
+class Testbed(toga.App):
+    def startup(self):
+        self.main_window = toga.MainWindow(title=self.formal_name)
+        self.main_window.content = toga.Box(
+            children=[
+                toga.Label("Did you forget to use --test?"),
+            ]
+        )
+        self.main_window.show()
+
+        # FIXME: workaround for https://github.com/beeware/rubicon-objc/issues/228
+        if toga.platform.current_platform == "iOS":
+            import asyncio
+
+            async def heartbeat(*args, **kwargs):
+                while True:
+                    await asyncio.sleep(0.0001)
+
+            self.add_background_task(heartbeat)
+        # END FIXME
+
+
+def main():
+    return Testbed(app_name="testbed")
diff --git a/testbed/src/testbed/resources/__init__.py b/testbed/src/testbed/resources/__init__.py
new file mode 100644
index 0000000000..e69de29bb2
diff --git a/testbed/src/testbed/resources/testbed-1024.png b/testbed/src/testbed/resources/testbed-1024.png
new file mode 100644
index 0000000000..d0ece53f8d
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-1024.png differ
diff --git a/testbed/src/testbed/resources/testbed-120.png b/testbed/src/testbed/resources/testbed-120.png
new file mode 100644
index 0000000000..fe725aaaf7
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-120.png differ
diff --git a/testbed/src/testbed/resources/testbed-152.png b/testbed/src/testbed/resources/testbed-152.png
new file mode 100644
index 0000000000..3293486604
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-152.png differ
diff --git a/testbed/src/testbed/resources/testbed-167.png b/testbed/src/testbed/resources/testbed-167.png
new file mode 100644
index 0000000000..48046cdd62
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-167.png differ
diff --git a/testbed/src/testbed/resources/testbed-180.png b/testbed/src/testbed/resources/testbed-180.png
new file mode 100644
index 0000000000..dd1233d701
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-180.png differ
diff --git a/testbed/src/testbed/resources/testbed-20.png b/testbed/src/testbed/resources/testbed-20.png
new file mode 100644
index 0000000000..35d1383fcb
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-20.png differ
diff --git a/testbed/src/testbed/resources/testbed-29.png b/testbed/src/testbed/resources/testbed-29.png
new file mode 100644
index 0000000000..b199573108
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-29.png differ
diff --git a/testbed/src/testbed/resources/testbed-40.png b/testbed/src/testbed/resources/testbed-40.png
new file mode 100644
index 0000000000..248f415f1a
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-40.png differ
diff --git a/testbed/src/testbed/resources/testbed-58.png b/testbed/src/testbed/resources/testbed-58.png
new file mode 100644
index 0000000000..aaad2800ad
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-58.png differ
diff --git a/testbed/src/testbed/resources/testbed-60.png b/testbed/src/testbed/resources/testbed-60.png
new file mode 100644
index 0000000000..686ae8f49a
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-60.png differ
diff --git a/testbed/src/testbed/resources/testbed-76.png b/testbed/src/testbed/resources/testbed-76.png
new file mode 100644
index 0000000000..2a6ca25939
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-76.png differ
diff --git a/testbed/src/testbed/resources/testbed-80.png b/testbed/src/testbed/resources/testbed-80.png
new file mode 100644
index 0000000000..f25442602a
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-80.png differ
diff --git a/testbed/src/testbed/resources/testbed-87.png b/testbed/src/testbed/resources/testbed-87.png
new file mode 100644
index 0000000000..af46ea995f
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-87.png differ
diff --git a/testbed/src/testbed/resources/testbed-round-144.png b/testbed/src/testbed/resources/testbed-round-144.png
new file mode 100644
index 0000000000..99bdd5abd6
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-round-144.png differ
diff --git a/testbed/src/testbed/resources/testbed-round-192.png b/testbed/src/testbed/resources/testbed-round-192.png
new file mode 100644
index 0000000000..2c69657d04
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-round-192.png differ
diff --git a/testbed/src/testbed/resources/testbed-round-48.png b/testbed/src/testbed/resources/testbed-round-48.png
new file mode 100644
index 0000000000..aa27b58e52
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-round-48.png differ
diff --git a/testbed/src/testbed/resources/testbed-round-72.png b/testbed/src/testbed/resources/testbed-round-72.png
new file mode 100644
index 0000000000..8489be821e
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-round-72.png differ
diff --git a/testbed/src/testbed/resources/testbed-round-96.png b/testbed/src/testbed/resources/testbed-round-96.png
new file mode 100644
index 0000000000..effcd8e391
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-round-96.png differ
diff --git a/testbed/src/testbed/resources/testbed-square-144.png b/testbed/src/testbed/resources/testbed-square-144.png
new file mode 100644
index 0000000000..3577821858
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-square-144.png differ
diff --git a/testbed/src/testbed/resources/testbed-square-192.png b/testbed/src/testbed/resources/testbed-square-192.png
new file mode 100644
index 0000000000..1a254076e8
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-square-192.png differ
diff --git a/testbed/src/testbed/resources/testbed-square-48.png b/testbed/src/testbed/resources/testbed-square-48.png
new file mode 100644
index 0000000000..70689ed130
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-square-48.png differ
diff --git a/testbed/src/testbed/resources/testbed-square-72.png b/testbed/src/testbed/resources/testbed-square-72.png
new file mode 100644
index 0000000000..509ce382f0
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-square-72.png differ
diff --git a/testbed/src/testbed/resources/testbed-square-96.png b/testbed/src/testbed/resources/testbed-square-96.png
new file mode 100644
index 0000000000..a1489805e0
Binary files /dev/null and b/testbed/src/testbed/resources/testbed-square-96.png differ
diff --git a/testbed/src/testbed/resources/testbed.icns b/testbed/src/testbed/resources/testbed.icns
new file mode 100644
index 0000000000..64cac52111
Binary files /dev/null and b/testbed/src/testbed/resources/testbed.icns differ
diff --git a/testbed/src/testbed/resources/testbed.ico b/testbed/src/testbed/resources/testbed.ico
new file mode 100644
index 0000000000..aa3492f61b
Binary files /dev/null and b/testbed/src/testbed/resources/testbed.ico differ
diff --git a/testbed/src/testbed/resources/testbed.png b/testbed/src/testbed/resources/testbed.png
new file mode 100644
index 0000000000..d52899e4c8
Binary files /dev/null and b/testbed/src/testbed/resources/testbed.png differ
diff --git a/testbed/tests/__init__.py b/testbed/tests/__init__.py
new file mode 100644
index 0000000000..e69de29bb2
diff --git a/testbed/tests/assertions.py b/testbed/tests/assertions.py
new file mode 100644
index 0000000000..bea59e6f1f
--- /dev/null
+++ b/testbed/tests/assertions.py
@@ -0,0 +1,16 @@
+from pytest import approx
+
+
+# This could be generalized in future to accept syntax like:
+#   * assert_set_get(obj, name, pytest.approx(value)) - for floating point values
+#   * assert_set_get(obj, name, set_value, get_value) - where the two values are different
+def assert_set_get(obj, name, value):
+    """Calls a setter, then asserts that the same value is returned by the getter."""
+    setattr(obj, name, value)
+    assert getattr(obj, name) == value
+
+
+def assert_color(actual, expected):
+    for component in ["r", "g", "b"]:
+        assert getattr(actual, component) == getattr(expected, component)
+    assert actual.a == approx(expected.a, abs=(1 / 255))
diff --git a/testbed/tests/conftest.py b/testbed/tests/conftest.py
new file mode 100644
index 0000000000..4808a614c1
--- /dev/null
+++ b/testbed/tests/conftest.py
@@ -0,0 +1,57 @@
+import asyncio
+import inspect
+from dataclasses import dataclass
+
+from pytest import fixture
+
+import toga
+
+
+@fixture(scope="session")
+def app():
+    return toga.App.app
+
+
+@fixture(scope="session")
+def main_window(app):
+    return app.main_window
+
+
+# Controls the event loop used by pytest-asyncio.
+@fixture(scope="session")
+def event_loop(app):
+    return ProxyEventLoop(app._impl.loop)
+
+
+# Proxy which forwards all tasks to another event loop in a thread-safe manner. It
+# implements only the methods used by pytest-asyncio.
+@dataclass
+class ProxyEventLoop(asyncio.AbstractEventLoop):
+    loop: object
+
+    # Used by ensure_future.
+    def create_task(self, coro):
+        return ProxyTask(coro)
+
+    def run_until_complete(self, future):
+        if inspect.iscoroutine(future):
+            coro = future
+        elif isinstance(future, ProxyTask):
+            coro = future.coro
+        else:
+            raise TypeError(f"Future type {type(future)} is not currently supported")
+        return asyncio.run_coroutine_threadsafe(coro, self.loop).result()
+
+    def close(self):
+        pass
+
+
+@dataclass
+class ProxyTask:
+    coro: object
+
+    # Used by ensure_future.
+    _source_traceback = None
+
+    def done(self):
+        return False
diff --git a/testbed/tests/data.py b/testbed/tests/data.py
new file mode 100644
index 0000000000..a3a52cf3d9
--- /dev/null
+++ b/testbed/tests/data.py
@@ -0,0 +1,25 @@
+from toga.colors import rgba
+
+# TODO: add non-ASCII strings.
+TEXTS = ["", " ", "a", "ab", "abc", "hello world", "hello\nworld"]
+
+
+# TODO: include None
+COLORS = [
+    rgba(r, g, b, a)
+    for r, g, b in [
+        # Black, gray, white,
+        (0, 0, 0),
+        (1, 1, 1),
+        (10, 10, 10),
+        (128, 128, 128),
+        (245, 245, 245),
+        (254, 254, 254),
+        (255, 255, 255),
+        # Primaries
+        (255, 0, 0),
+        (0, 255, 0),
+        (0, 0, 255),
+    ]
+    for a in [0.0, 0.01, 0.1, 0.5, 0.9, 0.99, 1.0]
+]
diff --git a/testbed/tests/testbed.py b/testbed/tests/testbed.py
new file mode 100644
index 0000000000..20c9c1b1b9
--- /dev/null
+++ b/testbed/tests/testbed.py
@@ -0,0 +1,61 @@
+import os
+import sys
+import tempfile
+from functools import partial
+from pathlib import Path
+from threading import Thread
+
+import pytest
+
+from testbed.app import main
+
+
+def run_tests(app):
+    project_path = Path(__file__).parent.parent
+    os.chdir(project_path)
+
+    # TODO: replace with extractPackages.
+    if hasattr(sys, "getandroidapilevel"):
+        import tests
+
+        chaquopy_extract_package(tests)
+
+    pytest.main(
+        [
+            # Output formatting
+            "-vv",
+            "--no-header",
+            "--tb=native",
+            "-rP",  # Show stdout from all tests, even if they passed.
+            "--color=no",
+            # Run all async tests and fixtures using pytest-asyncio.
+            "--asyncio-mode=auto",
+            # Override the cache directory to be somewhere known writable
+            "-o",
+            f"cache_dir={tempfile.gettempdir()}/.pytest_cache",
+            project_path / "tests",
+        ]
+    )
+    app.add_background_task(lambda app, **kwargs: app.exit())
+
+
+def chaquopy_extract_package(pkg):
+    finder = pkg.__loader__.finder
+    for path in pkg.__path__:
+        chaquopy_extract_dir(finder, finder.zip_path(path))
+
+
+def chaquopy_extract_dir(finder, zip_dir):
+    for filename in finder.listdir(zip_dir):
+        zip_path = f"{zip_dir}/{filename}"
+        if finder.isdir(zip_path):
+            chaquopy_extract_dir(finder, zip_path)
+        else:
+            finder.extract_if_changed(zip_path)
+
+
+if __name__ == "__main__":
+    app = main()
+    thread = Thread(target=partial(run_tests, app))
+    app.add_background_task(lambda app, *kwargs: thread.start())
+    app.main_loop()
diff --git a/testbed/tests/widgets/__init__.py b/testbed/tests/widgets/__init__.py
new file mode 100644
index 0000000000..e69de29bb2
diff --git a/testbed/tests/widgets/conftest.py b/testbed/tests/widgets/conftest.py
new file mode 100644
index 0000000000..9390336efb
--- /dev/null
+++ b/testbed/tests/widgets/conftest.py
@@ -0,0 +1,22 @@
+from pytest import fixture
+
+import toga
+
+from .probe import get_probe
+
+
+@fixture
+async def widget():
+    raise NotImplementedError("test modules must define a `widget` fixture")
+
+
+@fixture
+async def probe(main_window, widget):
+    box = toga.Box(children=[widget])
+    main_window.content = box
+
+    probe = get_probe(widget)
+    probe.assert_container(box)
+    yield probe
+
+    main_window.content = toga.Box()
diff --git a/testbed/tests/widgets/probe.py b/testbed/tests/widgets/probe.py
new file mode 100644
index 0000000000..a1cf52118a
--- /dev/null
+++ b/testbed/tests/widgets/probe.py
@@ -0,0 +1,12 @@
+from importlib import import_module
+
+from pytest import skip
+
+
+def get_probe(widget):
+    name = type(widget).__name__
+    try:
+        module = import_module(f"tests_backend.widgets.{name.lower()}")
+    except ModuleNotFoundError:
+        skip(f"No probe module for {name}")
+    return getattr(module, f"{name}Probe")(widget)
diff --git a/testbed/tests/widgets/properties.py b/testbed/tests/widgets/properties.py
new file mode 100644
index 0000000000..64e6f9117b
--- /dev/null
+++ b/testbed/tests/widgets/properties.py
@@ -0,0 +1,20 @@
+from ..assertions import assert_color, assert_set_get
+from ..data import COLORS, TEXTS
+
+
+async def test_text(widget, probe):
+    for text in TEXTS:
+        assert_set_get(widget, "text", text)
+        assert probe.text == text
+
+
+async def test_color(widget, probe):
+    for color in COLORS:
+        widget.style.color = color
+        assert_color(probe.color, color)
+
+
+async def test_background_color(widget, probe):
+    for color in COLORS:
+        widget.style.background_color = color
+        assert_color(probe.background_color, color)
diff --git a/testbed/tests/widgets/test_button.py b/testbed/tests/widgets/test_button.py
new file mode 100644
index 0000000000..6d3f0dff25
--- /dev/null
+++ b/testbed/tests/widgets/test_button.py
@@ -0,0 +1,26 @@
+from unittest.mock import Mock
+
+from pytest import fixture
+
+import toga
+
+from .properties import (  # noqa: F401
+    test_background_color,
+    test_color,
+    test_text,
+)
+
+
+@fixture
+async def widget():
+    return toga.Button("")
+
+
+async def test_press(widget, probe):
+    handler = Mock()
+    # TODO: can't use assert_set_get, because getattr returns the wrapped handler, which
+    # is an implementation detail that we shouldn't expose.
+    # https://github.com/beeware/toga/pull/804 may be relevant.
+    setattr(widget, "on_press", handler)
+    probe.press()
+    handler.assert_called_once_with(widget)
diff --git a/testbed/tests/widgets/test_label.py b/testbed/tests/widgets/test_label.py
new file mode 100644
index 0000000000..334e40faf9
--- /dev/null
+++ b/testbed/tests/widgets/test_label.py
@@ -0,0 +1,37 @@
+from pytest import approx, fixture, mark
+
+import toga
+
+from .properties import (  # noqa: F401
+    test_background_color,
+    test_color,
+    test_text,
+)
+
+
+@fixture
+async def widget():
+    return toga.Label("")
+
+
+# TODO: a `width` test, for any widget whose width depends on its text.
+@mark.skip("changing text does not trigger a refresh (#1289)")
+async def test_multiline(widget, probe):
+    def make_lines(n):
+        return "\n".join(f"line{i}" for i in range(n))
+
+    widget.text = make_lines(1)
+    # TODO: Android at least will need an `await` after each text change, to give the
+    # native layout a chance to update.
+    line_height = probe.height
+
+    widget.text = make_lines(2)
+    assert probe.height == approx(line_height * 2, rel=0.1)
+    line_spacing = probe.height - (line_height * 2)
+
+    for n in range(3, 10):
+        widget.text = make_lines(n)
+        assert probe.height == approx(
+            (line_height * n) + (line_spacing * (n - 1)),
+            rel=0.1,
+        )
diff --git a/testbed/tests/widgets/test_slider.py b/testbed/tests/widgets/test_slider.py
new file mode 100644
index 0000000000..44b40a67e6
--- /dev/null
+++ b/testbed/tests/widgets/test_slider.py
@@ -0,0 +1,73 @@
+from unittest.mock import Mock
+
+from pytest import approx, fixture, mark
+
+import toga
+from toga.platform import current_platform
+
+POSITIONS = [0, 0.01, 0.1, 0.5, 0.9, 0.99, 1]
+SCALES = [0.01, 0.1, 1, 10, 100000]
+
+# How accurate the position must be in continuous mode.
+ACCURACY = 0.001
+
+
+@fixture
+async def widget(on_change):
+    return toga.Slider(on_change=on_change)
+
+
+@fixture
+def on_change():
+    return Mock()
+
+
+@mark.skipif(current_platform == "android", reason="value is 0.0")
+async def test_init(widget, probe, on_change):
+    assert widget.value == 0.5
+    assert widget.range == (0, 1)
+    assert probe.position == 0.5
+    on_change.assert_not_called()
+
+
+@mark.skipif(current_platform == "windows", reason="on_change called 2 times")
+@mark.skipif(current_platform == "android", reason="position is 0.0")
+async def test_value(widget, probe, on_change):
+    for scale in SCALES:
+        widget.range = (0, scale)
+        for position in POSITIONS:
+            on_change.mock_calls.clear()
+            widget.value = position * scale
+            assert probe.position == approx(position, abs=ACCURACY)
+            on_change.assert_called_once_with(widget)
+
+
+@mark.skipif(
+    current_platform in ["android", "windows"], reason="on_change called 0 times"
+)
+async def test_change(widget, probe, on_change):
+    for scale in SCALES:
+        widget.range = (0, scale)
+        for position in POSITIONS:
+            on_change.mock_calls.clear()
+            probe.change(position)
+            assert widget.value == approx(position * scale, abs=(ACCURACY * scale))
+            on_change.assert_called_once_with(widget)
+
+
+@mark.skipif(current_platform == "windows", reason="value does not remain constant")
+@mark.skipif(current_platform == "android", reason="value is 0.0")
+async def test_min(widget, probe):
+    for min in POSITIONS[:4]:
+        widget.range = (min, 1)
+        assert widget.value == 0.5
+        assert probe.position == approx((0.5 - min) / (1 - min), abs=ACCURACY)
+
+
+@mark.skipif(current_platform == "windows", reason="value does not remain constant")
+@mark.skipif(current_platform == "android", reason="value is 0.0")
+async def test_max(widget, probe):
+    for max in POSITIONS[-4:]:
+        widget.range = (0, max)
+        assert widget.value == 0.5
+        assert probe.position == approx(0.5 / max, abs=ACCURACY)
diff --git a/tox.ini b/tox.ini
index a3da155e0d..3a89921100 100644
--- a/tox.ini
+++ b/tox.ini
@@ -54,9 +54,9 @@ commands =
 [testenv:package]
 skip_install = True
 deps =
-    check_manifest
-    build
-    twine
+    check_manifest==0.48
+    build==0.9.0
+    twine==4.0.2
 commands =
     check-manifest -v {posargs}
     python -m build {posargs}
diff --git a/winforms/MANIFEST.in b/winforms/MANIFEST.in
index 126da665ed..96861fad80 100644
--- a/winforms/MANIFEST.in
+++ b/winforms/MANIFEST.in
@@ -3,3 +3,4 @@ include LICENSE
 include README.rst
 include tox.ini
 recursive-include tests *.py
+recursive-include tests_backend *.py
diff --git a/winforms/tests_backend/__init__.py b/winforms/tests_backend/__init__.py
new file mode 100644
index 0000000000..e69de29bb2
diff --git a/winforms/tests_backend/widgets/__init__.py b/winforms/tests_backend/widgets/__init__.py
new file mode 100644
index 0000000000..e69de29bb2
diff --git a/winforms/tests_backend/widgets/base.py b/winforms/tests_backend/widgets/base.py
new file mode 100644
index 0000000000..5ab8afae78
--- /dev/null
+++ b/winforms/tests_backend/widgets/base.py
@@ -0,0 +1,44 @@
+from System import EventArgs, Object
+
+from .properties import toga_color
+
+
+class SimpleProbe:
+    def __init__(self, widget):
+        self.native = widget._impl.native
+        assert isinstance(self.native, self.native_class)
+
+    def assert_container(self, container):
+        container_native = container._impl.native
+        for control in container_native.Controls:
+            if Object.ReferenceEquals(control, self.native):
+                break
+        else:
+            raise ValueError(f"cannot find {self.native} in {container_native}")
+
+    @property
+    def enabled(self):
+        return self.native.Enabled
+
+    @property
+    def background_color(self):
+        return toga_color(self.native.BackColor)
+
+    @property
+    def color(self):
+        return toga_color(self.native.ForeColor)
+
+    @property
+    def hidden(self):
+        return not self.native.Visible
+
+    @property
+    def width(self):
+        return self.native.Width
+
+    @property
+    def height(self):
+        return self.native.Height
+
+    def press(self):
+        self.native.OnClick(EventArgs.Empty)
diff --git a/winforms/tests_backend/widgets/button.py b/winforms/tests_backend/widgets/button.py
new file mode 100644
index 0000000000..b04221b842
--- /dev/null
+++ b/winforms/tests_backend/widgets/button.py
@@ -0,0 +1,11 @@
+import System.Windows.Forms
+
+from .base import SimpleProbe
+
+
+class ButtonProbe(SimpleProbe):
+    native_class = System.Windows.Forms.Button
+
+    @property
+    def text(self):
+        return self.native.Text
diff --git a/winforms/tests_backend/widgets/label.py b/winforms/tests_backend/widgets/label.py
new file mode 100644
index 0000000000..f7ae735b8c
--- /dev/null
+++ b/winforms/tests_backend/widgets/label.py
@@ -0,0 +1,11 @@
+import System.Windows.Forms
+
+from .base import SimpleProbe
+
+
+class LabelProbe(SimpleProbe):
+    native_class = System.Windows.Forms.Label
+
+    @property
+    def text(self):
+        return self.native.Text
diff --git a/winforms/tests_backend/widgets/properties.py b/winforms/tests_backend/widgets/properties.py
new file mode 100644
index 0000000000..f0eba27506
--- /dev/null
+++ b/winforms/tests_backend/widgets/properties.py
@@ -0,0 +1,5 @@
+from toga.colors import rgba
+
+
+def toga_color(color):
+    return rgba(color.R, color.G, color.B, color.A / 255)
diff --git a/winforms/tests_backend/widgets/slider.py b/winforms/tests_backend/widgets/slider.py
new file mode 100644
index 0000000000..c1258159d8
--- /dev/null
+++ b/winforms/tests_backend/widgets/slider.py
@@ -0,0 +1,22 @@
+import System.Windows.Forms
+
+from .base import SimpleProbe
+
+
+class SliderProbe(SimpleProbe):
+    native_class = System.Windows.Forms.TrackBar
+
+    @property
+    def position(self):
+        return (self.native.Value - self._min) / (self._max - self._min)
+
+    def change(self, position):
+        self.native.Value = self._min + round(position * (self._max - self._min))
+
+    @property
+    def _min(self):
+        return self.native.Minimum
+
+    @property
+    def _max(self):
+        return self.native.Maximum