From: R. David Murray <rdmurray@bitdance.com>
Date: Mon, 28 Sep 2009 16:57:45 +0000 (+0000)
Subject: Prevent test_bad_address failure when a domain in the dns search
X-Git-Tag: v2.7a1~448
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=97f5ff37ffa259cbb8fa418be04b03eac0531251;p=python

Prevent test_bad_address failure when a domain in the dns search
path implements a '*' default rule.  Also update comment with
a more complete explanation of the difficulties inherent in
the test.
---

diff --git a/Lib/test/test_urllib2_localnet.py b/Lib/test/test_urllib2_localnet.py
index 1729d5f408..ee19a4b38e 100644
--- a/Lib/test/test_urllib2_localnet.py
+++ b/Lib/test/test_urllib2_localnet.py
@@ -472,14 +472,24 @@ class TestUrlopen(unittest.TestCase):
         # Make sure proper exception is raised when connecting to a bogus
         # address.
         self.assertRaises(IOError,
-                          # SF patch 809915:  In Sep 2003, VeriSign started
-                          # highjacking invalid .com and .net addresses to
-                          # boost traffic to their own site.  This test
-                          # started failing then.  One hopes the .invalid
-                          # domain will be spared to serve its defined
-                          # purpose.
-                          # urllib2.urlopen, "http://www.sadflkjsasadf.com/")
-                          urllib2.urlopen, "http://sadflkjsasf.i.nvali.d/")
+                          # Given that both VeriSign and various ISPs have in
+                          # the past or are presently hijacking various invalid
+                          # domain name requests in an attempt to boost traffic
+                          # to their own sites, finding a domain name to use
+                          # for this test is difficult.  RFC2606 leads one to
+                          # believe that '.invalid' should work, but experience
+                          # seemed to indicate otherwise.  Single character
+                          # TLDs are likely to remain invalid, so this seems to
+                          # be the best choice. The trailing '.' prevents a
+                          # related problem: The normal DNS resolver appends
+                          # the domain names from the search path if there is
+                          # no '.' the end and, and if one of those domains
+                          # implements a '*' rule a result is returned.
+                          # However, none of this will prevent the test from
+                          # failing if the ISP hijacks all invalid domain
+                          # requests.  The real solution would be to be able to
+                          # parameterize the framework with a mock resolver.
+                          urllib2.urlopen, "http://sadflkjsasf.i.nvali.d./")
 
 
 def test_main():