BUG: refguide check and CircleCI failures in master

Describe your issue.

There are two new CI failures in master that started today.

Error message

refguide (e.g. gh-15325)

File "doc/source/tutorial/interpolate.rst", line 261, in /home/vsts/work/1/s/tools/../doc/source/tutorial/interpolate.rst
Failed example:
    interpolate.sproot(tck)
Expected:
    array([3.1416])
Got:
    array([1.05630421e-17, 3.14159265e+00])
/home/vsts/work/1/s/tools/refguide_check.py:662: DeprecationWarning: elementwise comparison failed; this will raise an error in the future.
  if want == got:

CircleCI build_docs (e.g. gh-15326)

waiting for workers...
/home/circleci/repo/build/testenv/lib/python3.8/site-packages/scipy/signal/windows/_windows.py:docstring of scipy.signal.windows._windows.bohman:33: WARNING: Exception occurred in plotting scipy-signal-windows-bohman-1
 from /home/circleci/repo/doc/source/reference/generated/scipy.signal.windows.bohman.rst:
Traceback (most recent call last):
  File "/home/circleci/repo/venv/lib/python3.8/site-packages/matplotlib/sphinxext/plot_directive.py", line 517, in _run_code
    exec(code, ns)
  File "<string>", line 16, in <module>
RuntimeWarning: divide by zero encountered in log10
/home/circleci/repo/build/testenv/lib/python3.8/site-packages/scipy/signal/windows/_windows.py:docstring of scipy.signal.windows._windows.cosine:36: WARNING: Exception occurred in plotting scipy-signal-windows-cosine-1
 from /home/circleci/repo/doc/source/reference/generated/scipy.signal.windows.cosine.rst:
Traceback (most recent call last):
  File "/home/circleci/repo/venv/lib/python3.8/site-packages/matplotlib/sphinxext/plot_directive.py", line 517, in _run_code
    exec(code, ns)
  File "<string>", line 16, in <module>
RuntimeWarning: divide by zero encountered in log10

SciPy/NumPy/Python version information

master

[rgommers]

Copying the warnings here:

docstring of scipy.linalg.lapack.ssyevr:16: WARNING: Inline literal start-string without end-string.
docstring of scipy.linalg.lapack.ssyevr:22: WARNING: Inline literal start-string without end-string.
docstring of scipy.linalg.lapack.ssyevr:22: WARNING: Inline emphasis start-string without end-string.
docstring of scipy.linalg.lapack.ssyevr:22: WARNING: Inline emphasis start-string without end-string.
docstring of scipy.linalg.lapack.ssyevx:16: WARNING: Inline literal start-string without end-string.
docstring of scipy.linalg.lapack.ssyevx:22: WARNING: Inline literal start-string without end-string.
docstring of scipy.linalg.lapack.zheevr:16: WARNING: Inline literal start-string without end-string.
docstring of scipy.linalg.lapack.zheevx:16: WARNING: Inline literal start-string without end-string.
docstring of scipy.linalg.lapack.zheevx:16: WARNING: Inline emphasis start-string without end-string.
docstring of scipy.linalg.lapack.dsyevr:16: WARNING: Inline literal start-string without end-string.
docstring of scipy.linalg.lapack.dsyevr:22: WARNING: Inline literal start-string without end-string.
docstring of scipy.linalg.lapack.dsyevr:22: WARNING: Inline emphasis start-string without end-string.
docstring of scipy.linalg.lapack.dsyevr:22: WARNING: Inline emphasis start-string without end-string.
docstring of scipy.linalg.lapack.dsyevx:16: WARNING: Inline literal start-string without end-string.
docstring of scipy.linalg.lapack.dsyevx:22: WARNING: Inline literal start-string without end-string.
docstring of scipy.linalg.lapack.cheevr:16: WARNING: Inline literal start-string without end-string.
docstring of scipy.linalg.lapack.cheevx:16: WARNING: Inline literal start-string without end-string.
docstring of scipy.linalg.lapack.cheevx:16: WARNING: Inline emphasis start-string without end-string

These are coming from f2py-generated docstrings, which are ending up in build/scipy/linalg/_flapackmodule.c. There isn’t anything we can do to fix this I think, so let’s pin to numpy 1.21.5 in the doc build and see if it can be fixed in 1.22.1.

[rgommers]

The fix for the lapack issues will be in numpy 1.22.1.