[kaffe] Classloader bugs

[Daniel Bonniot]

Hi,

I think I am experimenting two Classloader bugs, in Kaffe CVS (built a 
few minutes ago). The good thing is that I have pretty small testcases.

Bug.java simply exercises loading from a URLClassLoader. The catch is 
that it is fragile on the presence or absence of a '/' at the end of the 
URL, for a file URL (this might also be a bug in java.net.URL, I am not 
sure).

(all this is on a Linux host)

$ javac Bug.java
$ java Bug $PWD Bug
$ kaffe Bug $PWD Bug
java.lang.ClassNotFoundException: Bug not found in 
[file:/home/daniel/tmp/kaffe-bug]
   at java.net.URLClassLoader.findClass (URLClassLoader.java:769)
   at java.lang.ClassLoader.loadClass (ClassLoader.java:142)
   at java.lang.ClassLoader.loadClass (ClassLoader.java:121)
   at Bug.main (Bug.java:9)
$ echo $PWD
/home/daniel/tmp/kaffe-bug
$ kaffe Bug $PWD/ Bug
$

Note how adding '/' at the end of the URL makes the classloader find the 
class.



The second bug is only slightly more complex. It involves a custom class 
loader. I have no interpretation of this bug at the moment, looks like 
an internal error.

$ javac Bug2.java
$ java Bug2 $PWD/ Bug2
$ kaffe Bug2 $PWD/ Bug2
java.lang.NullPointerException
   at java.lang.ClassLoader.defineClass0 (ClassLoader.java)
   at java.lang.ClassLoader.defineClass (ClassLoader.java:179)
   at java.security.SecureClassLoader.defineClass 
(SecureClassLoader.java:33)
   at java.net.URLClassLoader.findClass (URLClassLoader.java:855)
   at Bug2$1.loadClass (Bug2.java:23)
   at java.lang.ClassLoader.loadClass (ClassLoader.java:121)
   at Bug2.main (Bug2.java:36)

It's also interesting to see what happens when trying to load a class 
that does not exist:

$ java Bug2 $PWD/ Foo
Exception in thread "main" java.lang.ClassNotFoundException: Foo
        at java.net.URLClassLoader$1.run(URLClassLoader.java:198)
        at java.security.AccessController.doPrivileged(Native Method)
        at java.net.URLClassLoader.findClass(URLClassLoader.java:186)
        at java.lang.ClassLoader.loadClass(ClassLoader.java:299)
        at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:265)
        at java.lang.ClassLoader.loadClass(ClassLoader.java:255)
        at Bug2$1.loadClass(Bug2.java:28)
        at java.lang.ClassLoader.loadClass(ClassLoader.java:255)
        at Bug2.main(Bug2.java:36)
$ kaffe Bug2 $PWD/ Foo
java.lang.NullPointerException
   at Bug2$1.loadClass (Bug2.java:28)
   at java.lang.ClassLoader.loadClass (ClassLoader.java:121)
   at Bug2.main (Bug2.java:36)

I just checked the javadoc of java.lang.ClassLoader.getParent. It 
explicitely says that some implementations might return null to 
represent the bootstrap class loader. So the behaviour of kaffe in the 
second case (while searching for Foo) is not incorrect. It might still 
explain the bug in the first case (searching for Bug2) -- or not, since 
the lookup should succeed in the current classloader, not its parent. In 
any case, it sounds like a better idea to me to return the bootstrap 
class loader than to return null.

Cheers,

Daniel

-------------- next part --------------
A non-text attachment was scrubbed...
Name: Bug.java
Type: text/x-java
Size: 312 bytes
Desc: not available
Url : http://kaffe.org/pipermail/kaffe/attachments/20031117/be1cdb4b/attachment-0004.java 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Bug2.java
Type: text/x-java
Size: 1022 bytes
Desc: not available
Url : http://kaffe.org/pipermail/kaffe/attachments/20031117/be1cdb4b/attachment-0005.java