[Zope3-checkins] SVN: Zope3/trunk/src/zope/testbrowser/ Fix some bugs in mechanize.

[Stephan Richter]

Log message for revision 39691:
  Fix some bugs in mechanize.
  
  

Changed:
  A   Zope3/trunk/src/zope/testbrowser/BUGFIXES.txt
  U   Zope3/trunk/src/zope/testbrowser/mechanize/_mechanize.py

-=-
Added: Zope3/trunk/src/zope/testbrowser/BUGFIXES.txt
===================================================================
--- Zope3/trunk/src/zope/testbrowser/BUGFIXES.txt	2005-10-28 14:44:39 UTC (rev 39690)
+++ Zope3/trunk/src/zope/testbrowser/BUGFIXES.txt	2005-10-28 16:20:57 UTC (rev 39691)
@@ -0,0 +1,31 @@
+===================
+Mechanize Bug Fixes
+===================
+
+This file contains bug fixes to the mechanize framework that have not yet been
+reported to mechanize authors.
+
+``mechanize``
+-------------
+
+- The ``Browser.close()`` method fails to close the current response using
+  ``Browser._response.close()``. If you are opening a lot of pages, then you
+  will end up witha "too many files opnened" error eventually.
+
+- The ``update_history`` argument of ``Browser._mech_open()`` was unused.
+
+- In the ``Browser.close()`` method, the ``Browser._history`` attribute is set
+  to ``None``, but is should be ``[]`` (an empty list).
+
+- Related to the entry above, when the history is reset, none of the stored
+  responses is closed.
+
+- Due to the poor design of the history implementation, you can still get "too
+  many files opened" if you use the rest of the ``Browser`` API. The ideal
+  solution would be to close the previous response just before you use
+  ``Browser._mech_open()``. In fact this method should be responsible of the
+  cleanup.
+
+- URLs that are retrieved using pullparser might be unclean, meaning they
+  might contain leading and trailing whitespace. The ``Link`` class should
+  thus strip the base and link url.


Property changes on: Zope3/trunk/src/zope/testbrowser/BUGFIXES.txt
___________________________________________________________________
Name: svn:eol-style
   + native

Modified: Zope3/trunk/src/zope/testbrowser/mechanize/_mechanize.py
===================================================================
--- Zope3/trunk/src/zope/testbrowser/mechanize/_mechanize.py	2005-10-28 14:44:39 UTC (rev 39690)
+++ Zope3/trunk/src/zope/testbrowser/mechanize/_mechanize.py	2005-10-28 16:20:57 UTC (rev 39691)
@@ -51,6 +51,8 @@
 class Link:
     def __init__(self, base_url, url, text, tag, attrs):
         assert None not in [url, tag, attrs]
+        url = url.strip()
+        base_url = base_url.strip()
         self.base_url = base_url
         self.absolute_url = urlparse.urljoin(base_url, url)
         self.url, self.text, self.tag, self.attrs = url, text, tag, attrs
@@ -96,17 +98,23 @@
         self._history = []  # LIFO
         self.request = self._response = None
         self.form = None
-        self._forms = None
-        self._title = None
-        self._links = None
+        self._forms = self._title = self._links = None
         UserAgent.__init__(self)  # do this last to avoid __getattr__ problems
 
     def close(self):
         UserAgent.close(self)
-        self._history = self._forms = self._title = self._links = None
+        if self._response:
+            self._response.close()
         self.request = self._response = None
+        self.form = None
+        [response.close()
+         for request, response in self._history
+         if response is not None]
+        self._history = []
+        self._forms = self._title = self._links = None
 
-    def open(self, url, data=None): return self._mech_open(url, data)
+    def open(self, url, data=None):
+        return self._mech_open(url, data)
 
     def _mech_open(self, url, data=None, update_history=True):
         if not hasattr(url, 'get_full_url'):
@@ -120,8 +128,6 @@
                         "can't fetch relative URL: not viewing any document")
                 url = urlparse.urljoin(self._response.geturl(), url)
 
-        if self.request is not None:
-            self._history.append((self.request, self._response))
         self._response = None
         # we want self.request to be assigned even if OpenerDirector.open fails
         self.request = self._request(url, data)
@@ -133,6 +139,9 @@
             self._response = response_seek_wrapper(self._response)
         self._parse_html(self._response)
 
+        if (self.request is not None) and update_history:
+            self._history.append((self.request, self._response))
+
         return self._response
 
     def response(self):