jython-dev

[Jython-dev] jreload--it IS possible to reload a class in the same classloader

[Updike, C. <Cla...@jh...>]

regarding http://www.jython.org/docs/jreload.html and the statement:

"Why there is no support for reloading a single class? Java classes are
loaded through class-loaders, actually there is no support in java to
redefine an already loaded class through its class-loader."

I don't mean to split hairs here, but it IS actually possible to redefine an
already-loaded class in the same classloader as long as you use Class
methods to load it.  i.e. 
use:

Class redefineable = Class.forName("MyRedefinealbeClass");
redefineable.newInstance();

// Assume the previous version of MyRedefineableClass.class was modified
// Reload the new one...
redefineable  = null;
System.gc();
Class redefineable = Class.forName("MyRedefineableClass");
redefineable.newInstance();

Instead of:

MyUNRedefinableClass unRedefineable = new MyUNRedefineableClass();
murdc = null;

// does NOT pick up any changes to MyRedefineableClass.class
unRedefineable = new MyUNRedefineableClass(); 


This functionality is described in the following article,
http://www.sys-con.com/java/articleprint.cfm?id=307 

I don't know if this adds any new possibilities for jreload or not (or if
I'm wrong), but thought it worth pointing out anyway.  Note that the article
makes a call to System.gc() to "force" the classloader to dump the current
version but I don't believe System.gc() guarantees that this will happen.

-Clark

Re: [Jython-dev] jreload--it IS possible to reload a class in the same classloader

[Samuele P. <ped...@bl...>]

Have you tried that. Has he tried that?

I have tried with:

1)
java version "1.3.0"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.3.0-C)
Java HotSpot(TM) Client VM (build 1.3.0-C, mixed mode)

2)
java version "1.1.7A"

3)
Caricatore riga comandi Microsoft (R) per Java Versione 5.00.3802
Copyright (C) Microsoft Corp 1996-2000. Tutti i diritti riservati.
(It's jview, Versione is version)

Funny, Funny.
(To be honest I didn't need to try, jview not with standing)

regards, Samuele Pedroni.

----- Original Message -----
From: Updike, Clark <Cla...@jh...>
To: <jyt...@li...>
Sent: Monday, January 28, 2002 4:15 PM
Subject: [Jython-dev] jreload--it IS possible to reload a class in the same
classloader


> regarding http://www.jython.org/docs/jreload.html and the statement:
>
> "Why there is no support for reloading a single class? Java classes are
> loaded through class-loaders, actually there is no support in java to
> redefine an already loaded class through its class-loader."
>
> I don't mean to split hairs here, but it IS actually possible to redefine an
> already-loaded class in the same classloader as long as you use Class
> methods to load it.  i.e.
> use:
>
> Class redefineable = Class.forName("MyRedefinealbeClass");
> redefineable.newInstance();
>
> // Assume the previous version of MyRedefineableClass.class was modified
> // Reload the new one...
> redefineable  = null;
> System.gc();
> Class redefineable = Class.forName("MyRedefineableClass");
> redefineable.newInstance();
>
> Instead of:
>
> MyUNRedefinableClass unRedefineable = new MyUNRedefineableClass();
> murdc = null;
>
> // does NOT pick up any changes to MyRedefineableClass.class
> unRedefineable = new MyUNRedefineableClass();
>
>
> This functionality is described in the following article,
> http://www.sys-con.com/java/articleprint.cfm?id=307
>
> I don't know if this adds any new possibilities for jreload or not (or if
> I'm wrong), but thought it worth pointing out anyway.  Note that the article
> makes a call to System.gc() to "force" the classloader to dump the current
> version but I don't believe System.gc() guarantees that this will happen.
>
> -Clark
>
> _______________________________________________
> Jython-dev mailing list
> Jyt...@li...
> https://lists.sourceforge.net/lists/listinfo/jython-dev
>

Re: [Jython-dev] jreload--it IS possible to reload a class in the same classloader

[<bc...@wo...>]

[Samuele]

>... Has he tried that?

Well, the listing 2 example didn't even compile.

  http://www.sys-con.com/java/source.cfm?id=307

But even if he did try and and he found a VM that supported it, the
behavior he describe is in clear and direct conflict with the JLS:

: http://java.sun.com/docs/books/jls/second_edition/html/execution.doc.html#74294

It is rare to see the JLS spend that much text on stating a point.


>I have tried with:

And I tried with IBMJava130:

  java version "1.3.0"
  Java(TM) 2 Runtime Environment, Standard Edition (build 1.3.0)
  Classic VM (build 1.3.0, J2RE 1.3.0 IBM build cn130-20001018 (JIT
  disabled))

It too only loads the Bar class once, as it should.

regards,
finn

Re: [Jython-dev] jreload--it IS possible to reload a class in the same classloader

[Samuele P. <ped...@bl...>]

[oops, Finn has taken over my java classloaders guy role <wink>]

[myself]
> Have you tried that. Has he tried that?
>
> I have tried with:
>
> 1)
> java version "1.3.0"
> Java(TM) 2 Runtime Environment, Standard Edition (build 1.3.0-C)
> Java HotSpot(TM) Client VM (build 1.3.0-C, mixed mode)

Does not work.


> 2)
> java version "1.1.7A"

Does work (see later).

> 3)
> Caricatore riga comandi Microsoft (R) per Java Versione 5.00.3802
> Copyright (C) Microsoft Corp 1996-2000. Tutti i diritti riservati.
> (It's jview, Versione is version)

Does not work.

> Funny, Funny.
> (To be honest I didn't need to try, jview not with standing)
>
Ok, I didn't know/remember about 2) but see later.


4) various sun jvm on solaris:
java version "1.1.8"
java version "1.1.6"

java version "1.3.1"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.3.1-b24)
Java HotSpot(TM) Client VM (build 1.3.1-b24, mixed mode)

Does not work.

The 2nd edition of the JLS states a clear rule about class unloading
(btw class unloading is not mandatory but Jython works only on
jvms that implement it following the JLS rule). The rule is the one cited by
Finn:

http://java.sun.com/docs/books/jls/second_edition/html/execution.doc.html#74294

Btw the rule appeared in a set of Clarifications and Amendments to
The Java Language Specification (1st) edition:
http://java.sun.com/docs/books/jls/clarify.html

The clarifition about class unloading targets:
http://java.sun.com/docs/books/jls/first_edition/html/12.doc.html#44850

So: what he construes as > java 1.1 a feature:
- is a bug for a post clarication jvm (and indeed buggy jvms exist <wink>)
- a possible implementation feature for older jvms, but one portable code
cannot count on

So the guy should have known better (because even if at the time that was an
undocumented
feature it should have said whether he was talking about thin ice),

and we try our best to make informed and portable design decision.

[Btw java 1.4 offers class redefining features but through the debugging
interface
(not something that for  we can exploit as it is)
http://java.sun.com/j2se/1.4/docs/guide/jpda/enhancements.html#hotswap
]

regards, Samuele Pedroni.