TST: Add future dependency tests as a weekly CI job

Test future numpy and pandas versions with Matplotlib to see if
anything needs to be done in the future to address deprecations
or other pending changes.

Turns all warnings into errors, but filters out the distutils
deprecation and find_spec warnings.

Author: greglucas

.github/workflows/tests.yml

@@ -8,6 +8,9 @@ on:
   pull_request:
     branches-ignore:
       - v[0-9]+.[0-9]+.[0-9x]+-doc
+  schedule:
+    # 3:47 UTC on Saturdays
+    - cron: "47 3 * * 6"
 
 env:
   NO_AT_BRIDGE: 1  # Necessary for GTK3 interactive test.
@@ -209,6 +212,20 @@ jobs:
             echo 'wxPython is available' ||
             echo 'wxPython is not available'
 
+      - name: Install the nightly dependencies
+        # Only install the nightly dependencies during the scheduled event
+        if: ${{ github.event_name == 'schedule' && matrix.name-suffix != '(Minimum Versions)' }}
+        run: |
+          python -m pip install -i https://pypi.anaconda.org/scipy-wheels-nightly/simple --upgrade numpy pandas
+
+          # Turn all warnings to errors, except ignore the distutils deprecations and the find_spec warning
+          cat >> pytest.ini << EOF
+          filterwarnings =
+              error
+              ignore:.*distutils:DeprecationWarning
+              ignore:DynamicImporter.find_spec\(\) not found; falling back to find_module\(\):ImportWarning
+          EOF
+
       - name: Install Matplotlib
         run: |
           ccache -s
@@ -266,3 +283,18 @@ jobs:
         with:
           name: "${{ matrix.python-version }} ${{ matrix.os }} ${{ matrix.name-suffix }} result images"
           path: ./result_images
+
+      - name: Create issue on failure
+        uses: imjohnbo/issue-bot@v3
+        if: ${{ failure() && github.event_name == 'schedule' }}
+        with:
+          title: "[TST] Upcoming dependency test failures"
+          body: |
+            The weekly build with nightly wheels from numpy and pandas
+            has failed. Check the logs for any updates that need to be
+            made in matplotlib.
+
+          pinned: false
+          close-previous: false
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

lib/matplotlib/tests/test_axes.py

@@ -3438,6 +3438,8 @@ def test_errorbar():
     ax.errorbar(x, y, yerr=[yerr_lower, 2*yerr], xerr=xerr,
                 fmt='o', ecolor='g', capthick=2)
     ax.set_title('Mixed sym., log y')
+    # Force limits due to floating point slop potentially expanding the range
+    ax.set_ylim(1e-2, 1e1)
 
     fig.suptitle('Variable errorbars')
 

lib/matplotlib/tests/test_colors.py

@@ -568,14 +568,15 @@ def test_Normalize():
     # Don't lose precision on longdoubles (float128 on Linux):
     # for array inputs...
     vals = np.array([1.2345678901, 9.8765432109], dtype=np.longdouble)
-    norm = mcolors.Normalize(vals.min(), vals.max())
-    assert_array_equal(np.asarray(norm(vals)), [0, 1])
+    norm = mcolors.Normalize(vals[0], vals[1])
+    assert norm(vals).dtype == np.longdouble
+    assert_array_equal(norm(vals), [0, 1])
     # and for scalar ones.
     eps = np.finfo(np.longdouble).resolution
     norm = plt.Normalize(1, 1 + 100 * eps)
     # This returns exactly 0.5 when longdouble is extended precision (80-bit),
     # but only a value close to it when it is quadruple precision (128-bit).
-    assert 0 < norm(1 + 50 * eps) < 1
+    np.testing.assert_array_almost_equal_nulp(norm(1 + 50 * eps), 0.5)
 
 
 def test_FuncNorm():