diff --git a/scrapy/trunk/scrapy/utils/url.py b/scrapy/trunk/scrapy/utils/url.py
index b989f1941..89a29d5e9 100644
--- a/scrapy/trunk/scrapy/utils/url.py
+++ b/scrapy/trunk/scrapy/utils/url.py
@@ -138,6 +138,7 @@ def canonicalize_url(url, keep_blank_values=False, keep_fragments=False):
     For examples see the tests in scrapy.tests.test_utils_url
     """
 
+    url = str(url) # URLs *must* be a string; otherwise urllib may fail.
     parts = list(urlparse.urlparse(url))
     keyvals = cgi.parse_qsl(parts[4], keep_blank_values)
     keyvals.sort()