On 31/08/2012 12:05, Konstantin Kolinko wrote:
> 2012/8/31  <ma...@apache.org>:
>> Author: markt
>> Date: Thu Aug 30 22:55:20 2012
>> New Revision: 1379206
>>
>> URL: http://svn.apache.org/viewvc?rev=1379206&view=rev
>> Log:
>> Fix https://issues.apache.org/bugzilla/show_bug.cgi?id=53801
>> Overlapping URL patterns were sometimes merged incorrectly in security 
>> constraints leading to incorrect 401 responses. Note: it was possible for 
>> access to be denied when it should have been granted but it was not possible 
>> for access to be granted when it should have been denied.
>>
>> Modified:
>>     tomcat/trunk/java/org/apache/catalina/realm/RealmBase.java
>>
>> Modified: tomcat/trunk/java/org/apache/catalina/realm/RealmBase.java
>> URL: 
>> http://svn.apache.org/viewvc/tomcat/trunk/java/org/apache/catalina/realm/RealmBase.java?rev=1379206&r1=1379205&r2=1379206&view=diff
>> ==============================================================================
>> --- tomcat/trunk/java/org/apache/catalina/realm/RealmBase.java (original)
>> +++ tomcat/trunk/java/org/apache/catalina/realm/RealmBase.java Thu Aug 30 
>> 22:55:20 2012
>> @@ -629,14 +629,15 @@ public abstract class RealmBase extends
>>                      }
>>                  }
>>                  if(matched) {
>> -                    found = true;
>>                      if(length > longest) {
>> +                        found = false;
>>                          if(results != null) {
>>                              results.clear();
>>                          }
>>                          longest = length;
>>                      }
>>                      if(collection[j].findMethod(method)) {
>> +                        found = true;
>>                          if(results == null) {
>>                              results = new ArrayList<>();
>>                          }
> 
> There are several loops over constraints, with
> [[[
>  if(found) {
>  return resultsToArray(results);
> }
> ]]]
> between them, and only one of such loops is fixed by this commit.

Correct. Only in one loop is there the possibility that a previously
found constraint will be removed. The bug was that the found flag was
not reset in this case and an empty set of constraints returned rather
than null.

Mark


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org

Reply via email to