commons-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Boris (JIRA)" <j...@apache.org>
Subject [jira] Reopened: (LANG-481) Possible race-conditions in hashCode of the range classes
Date Sat, 10 Oct 2009 19:26:31 GMT

     [ https://issues.apache.org/jira/browse/LANG-481?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Boris reopened LANG-481:
------------------------


I'm sorry, I just dicovered that there is another data-race in that methods that I overlooked
before.

In short: Due to the Java Memory Model and allowed reorderings it is possible that the hashCode()
can return 0 instead of the correct value and thus breaking HashMaps etc. in this case.
Please see [this post|http://jeremymanson.blogspot.com/2008/12/benign-data-races-in-java.html]
for details on this issue and String.hashCode() in the java-sources as a reference how to
do it right. He explained it better than I can.

I'll attach a patch against trunk to fix this.

> Possible race-conditions in hashCode of the range classes
> ---------------------------------------------------------
>
>                 Key: LANG-481
>                 URL: https://issues.apache.org/jira/browse/LANG-481
>             Project: Commons Lang
>          Issue Type: Bug
>    Affects Versions: 2.4
>            Reporter: Boris
>            Priority: Minor
>             Fix For: 3.0
>
>         Attachments: LANG-481.patch
>
>
> The hashCode() methods of the range classes look very suspicious to me. The value is
lazily initialized, but the calculation is done _on the cached value. With some unlucky timing
a caller may get an incomplete hash.
> An unlucky sequence of Code could be something like
> T1:        if (hashCode == 0) // true
> T1:            hashCode = 17;
> T2:         if (hashCode == 0) // now false because hashCode was already set to 17
> T2:         return hashCode; // return 17
> T1:            hashCode = 37 * hashCode...........
> where T1 and T2 are different threads accessing the method in parallel and T2 gets the
wrong hash "17".
> Affected classes are
> org.apache.commons.lang.math.DoubleRange
> org.apache.commons.lang.math.FloatRange
> org.apache.commons.lang.math.IntRange
> org.apache.commons.lang.math.LongRange
> org.apache.commons.lang.math.NumberRange
> org.apache.commons.lang.math.Range
> Possible fix: calculate the hash on a temporary variable and finally assign it to the
member

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message