activemq-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Hiram Chirino (JIRA)" <j...@apache.org>
Subject [jira] Commented: (AMQ-704) ActiveIO compromises java.lang.System.properties
Date Thu, 04 May 2006 04:45:41 GMT
    [ https://issues.apache.org/activemq/browse/AMQ-704?page=comments#action_36140 ] 

Hiram Chirino commented on AMQ-704:
-----------------------------------

The reason the singleton map was put in the  System properties were chosen in the first place
is because it's guaranteed to be in the primordial classloader and can be acessed by all classloaders.
 Putting a singleton in a static class field would now produce a system wide singleton if
multiple classloaders are in use for the class that is holding the static hashmap.

Do we know of any other locations that we can register a JVM wide object that is accessible
by all classloaders?

Stringifing the map sounds promising if a bit in-efficient.  But then again locking is not
a fast thing right?

For now, activeio has implemented a workaround.  Add a -Dorg.apache.activeio.journal.active.DisableLocking=true
to JVM startup options and activeio will avoid using the System properties this way.

> ActiveIO compromises java.lang.System.properties
> ------------------------------------------------
>
>          Key: AMQ-704
>          URL: https://issues.apache.org/activemq/browse/AMQ-704
>      Project: ActiveMQ
>         Type: Bug

>  Environment: ActiveMQ 4.0RC2 (confirmed still around in the current ActiveIO SVN), 
Java 1.5 on Windows (but should be a problem anywhere).
>     Reporter: matt hoffman
>      Fix For: 4.0 RC2

>
>
> When using the ActiveIO journal, it's putting a java.util.HashSet into the System.properties
(using a properties.put() call that circumvents the  normal System.putProperty() method, that
only allows strings).   This isn't allowed, according to java.util.Properties' contract, and
therefore breaks other packages that rely on System.properties adhering to its contract. 
Most notably, it chokes Hibernate; however, anything that iterates through the System properties
will fail after ActiveIO has compromised it with a non-String value.
> Honestly, I don't know why Properties doesn't allow non-String values, but oh well. 
From the java.lang.Properties docs: 
> "Because Properties inherits from Hashtable, the put and putAll methods can be applied
to a Properties object. Their use is strongly discouraged as they allow the caller to insert
entries whose keys or values are not Strings. The setProperty method should be used instead.
If the store or save method is called on a "compromised" Properties object that contains a
non-String key or value, the call will fail."
> The call is being made in org.apache.activeio.journal.active.ControlFile; I suppose a
hack could be created that stringifies the HashSet before placing it in the properties and
recreates it before getting it out again; atlernately, a simple singleton or static set could
be used.  I'd be happy to implement either one, if you'd like. 

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
   https://issues.apache.org/activemq/secure/Administrators.jspa
-
For more information on JIRA, see:
   http://www.atlassian.com/software/jira


Mime
View raw message