www.jugpadova.it: Tag bug

Tomcat 5.5.20 and JavaMail Sessions

Lucio Benfante — Tue, 13 Mar 2007 07:40:00 -0500

You’ll find an article in italian on the same topic in my personal blog.

In Tomcat you can define a JNDI Resource for a JavaMail session putting the following code in you Context definition:

<Resource name="mail/Session" auth="Container"
            type="javax.mail.Session"
            mail.smtp.host="localhost"/>

And in you web.xml:

<resource-ref>
  <res-ref-name>mail/Session</res-ref-name>
  <res-type>javax.mail.Session</res-type>
  <res-auth>Container</res-auth>
</resource-ref>

Then you can use it in you code:

Context initCtx = new InitialContext();
Session session =
    (Session) envCtx.lookup("java:comp/env/mail/Session");

Using Tomcat 5.5.20 (and 5.5.17, and 5.5.23…and maybe other versions) the last statement will produce a ClassNotFoundException: org.apache.naming.factory.MailSessionFactory. The cause is obviously that class is missing from the common/lib/naming-factory.jar distributed by Tomcat. I don’t know if it was a (repeated) problem in building Tomcat. Luckly the source code of that class is still present in the source distribution of Tomcat. Simply it isn’t included in the build process if you don’t have the JavaMail and the JavaBeans Activation Framework in the build classpath. So I re-built the naming-factory.jar file and substituted in common/lib.

You can rebuild your own JAR, or download my new naming-factory.jar.

I discovered this problem (bug?) installing LifeRay 4.1.2 using the WAR distribution. On the contrary the LifeRay-Tomcat bundle contains the correct naming-factory.jar yet. The LifeRay error in this case is more obscure. It reports a javax.naming.NameNotFoundException: Name mail is not bound in this Context! Of course the JNDI Resource is correctly configured. This happens because LifeRay try to guess the JNDI name of the resource, that changes for the different application servers. So that exception comes from one of the attempts, and the real problem is lost and hidden. So, if you’ll see a such message, don’t spend time re-re-re-configuring your JNDI resources…but patch Tomcat!

Bug storico...

Lucio Benfante — Fri, 16 Jun 2006 07:03:00 -0500

...presente nel JDK, e probabilmente nella maggior parte delle implementazioni di algoritmi con approccio “divide et impera”.

Quanti di noi non hanno mai scritto questa semplice e innocente riga di codice (con low e high di tipo int)?

int mid = (low+high)/2;

Il problema è che la somma dei due interi può risultare in un overflow…gosh…

L’implementazione corretta sarebbe qualcosa di simile a:

int mid = low + ((high-low)/2);

Provate a cercare in qualunque libro di algoritmi e vedrete che anche i grandi sbagliano. Io ho guardato in ‘The Art of Computer Programming’), e anche l’implementazione dell’algoritmo di ricerca binaria di Donald E. Knuth soffre dello stesso bug.

Per quanto riguarda il JDK, il bug si trova (almeno) nel metodo binarySearch della classe java.utils.Arrays. A me non è mai capitato, ma ve l’immaginate la sorpresa (per non dire altro) di vedervi sollevare un ArrayIndexOutOfBoundsException da un metodo di quella classe?

L’annuncio e la discussione del bug li fa lo stesso autore della classe, Josh Bloch, in questo blog.

findbugs to discover bugs

Enrico Giurin — Wed, 17 May 2006 12:15:00 -0500

I was investigating about a good tool for profiling when I met with findbugs. To be honest I found discussion about this tool in a thread of jug milano mailing list. findbugs is a free software which looks for bugs in Java code. In order to execute this program is enough you have installed on your machine Java Web Start. Thereby try to click here , after set parameter of your project (src, jar, etc) and finally push on ‘Find Bugs’ button.
I was amazed at how many bugs it found like comparison of String objects using == or != instead of using equals() method.
By the way…I didn’t generate that code :-)