phys_cursor's hpos is overwritten. This is still not completely
correct, as it doesn't really make sense to use hpos at all to
get the cursor glyph (as that is relative to the width of the
characters on the line, which may have changed during the update).
hpos, hpos + len,
DRAW_NORMAL_TEXT, 0);
+ /* Invalidate old phys cursor if the glyph at its hpos is redrawn. */
+ if (updated_area == TEXT_AREA
+ && updated_window->phys_cursor_on_p
+ && updated_window->phys_cursor.vpos == output_cursor.vpos
+ && updated_window->phys_cursor.hpos >= hpos
+ && updated_window->phys_cursor.hpos < hpos + len)
+ updated_window->phys_cursor_on_p = 0;
+
UNBLOCK_INPUT;
/* Advance the output cursor. */