Don't pass *markevery* to errorbar lines/caps.

QuLogic · QuLogic · commit cc9ae2fe7f96 · 2020-09-24T02:54:05.000-04:00
That argument is applied to the line, but errorbar masking is determined
by *errorevery* (as actual masks), so *markevery* should not be
propagated.
diff --git a/lib/matplotlib/axes/_axes.py b/lib/matplotlib/axes/_axes.py
@@ -3219,6 +3219,7 @@ def errorbar(self, x, y, yerr=None, xerr=None,
         base_style.pop('markerfacecolor', None)
         base_style.pop('markeredgewidth', None)
         base_style.pop('markeredgecolor', None)
+        base_style.pop('markevery', None)
         base_style.pop('linestyle', None)
 
         # Make the style dict for the line collections (the bars).
diff --git a/lib/matplotlib/tests/test_axes.py b/lib/matplotlib/tests/test_axes.py
@@ -3390,7 +3390,7 @@ def test_errorbar_with_prop_cycle(fig_test, fig_ref):
 
 
 @check_figures_equal()
-def test_errorbar_offsets(fig_test, fig_ref):
+def test_errorbar_every(fig_test, fig_ref):
     x = np.linspace(0, 1, 15)
     y = x * (1-x)
     yerr = y/6
@@ -3411,6 +3411,11 @@ def test_errorbar_offsets(fig_test, fig_ref):
         ax_ref.errorbar(x[shift::4], y[shift::4], yerr[shift::4],
                         capsize=4, c=color, fmt='none')
 
+    # Check that markevery is propagated to line, without affecting errorbars.
+    ax_test.errorbar(x, y + 0.1, yerr, markevery=(1, 4), capsize=4, fmt='o')
+    ax_ref.plot(x[1::4], y[1::4] + 0.1, 'o', zorder=2.1)
+    ax_ref.errorbar(x, y + 0.1, yerr, capsize=4, fmt='none')
+
 
 @image_comparison(['hist_stacked_stepfilled', 'hist_stacked_stepfilled'])
 def test_hist_stacked_stepfilled():
