fix: checking for internet connection can timeout (#1813)

* fix: checking for internet connection can timeout

In some CI environments or low speed connections, ``HAS_INTERNET`` will
return False due to a TimeoutError.

This fix speeds up the connection speed by avoiding a DNS lookup and
application layer connection through HTTP. ``urlopen`` will also return
a TimeoutError if the DNS lookup fails.

* docs: update docstring, argument name

* docs: update docstring

* docs: add entry to whats-new

Author: gampnico

docs/whats-new.rst

@@ -61,7 +61,8 @@ Enhancements
   ``workflow.merge_gridded_data``. If no grid is provided, the default is to
   merge all grids of the provided gdirs (:pull:`1779`).
   By `Patrick Schmitt <https://github.com/pat-schmitt>`_
-
+- Fixed ``TimeoutError`` when running tests when urlopen hangs on DNS lookup (:pull:`1813`).
+  By `Nicolas Gampierakis <https://github.com/gampnico>`_.
 Bug fixes
 ~~~~~~~~~
 

oggm/tests/__init__.py

@@ -1,9 +1,10 @@
 import os
-from packaging.version import Version
+import socket
+from urllib.request import URLError, urlopen
 
-import pytest
 import matplotlib.ft2font
-from urllib.request import urlopen, URLError
+import pytest
+from packaging.version import Version
 
 from oggm import cfg
 from oggm.utils import SAMPLE_DATA_COMMIT
@@ -12,18 +13,34 @@
 
 # Matplotlib version changes plots, too
 HAS_MPL_FOR_TESTS = False
-if Version(matplotlib.__version__) >= Version('2'):
+if Version(matplotlib.__version__) >= Version("2"):
     HAS_MPL_FOR_TESTS = True
     BASELINE_DIR = os.path.join(cfg.CACHE_DIR,
                                 'oggm-sample-data-%s' % SAMPLE_DATA_COMMIT,
                                 'baseline_images', 'freetype_28')
 
-# quick n dirty method to see if internet is on
-try:
-    _ = urlopen('http://www.google.com', timeout=1)
-    HAS_INTERNET = True
-except URLError:
-    HAS_INTERNET = False
+def check_internet_access(
+    hostname: str = "8.8.8.8", port: int = 53, timeout: int = 1
+):
+    """Check if Internet is available.
+    
+    hostname : str, default "8.8.8.8"
+        Web address. Can be a public DNS or an HTTP link.
+    port : int, default 53
+        An open and unfiltered port number. This should be 53 for
+        the domain, or 443 for https.
+    timeout : int, default 1
+        Time in seconds before the connection times out.
+
+    """
+    try:
+        socket.setdefaulttimeout(timeout)
+        socket.socket(socket.AF_INET, socket.SOCK_STREAM).connect((hostname, port))
+        return True
+    except socket.error as e:
+        return False
+
+HAS_INTERNET = check_internet_access()
 
 
 def mpl_image_compare(baseline_dir=BASELINE_DIR, tolerance=1, **kwargs):