aries-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "David Jencks (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ARIES-669) Proxy or weaving code should mark all its changes synthetic
Date Mon, 06 Jun 2011 22:12:59 GMT

    [ https://issues.apache.org/jira/browse/ARIES-669?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13045164#comment-13045164
] 

David Jencks commented on ARIES-669:
------------------------------------

Your comment doesn't make sense to me.  How would you add an existing static init method?
 Isn't it already there?

Having looked at the class file and jvm specs a little bit and guessed quite a bit I think
that:

each static code in the original source gets turned into a static_init_* method of some kind

each class gets a <clinit> method that calls the static_init* methods in the order they
appear in the original source code.  This is not supposed to be marked synthetic.

I don't understand where the code from 

static int foo = 1;

goes.

I think that you want aries to add another static_init_foo method to do something, and at
this point I think you agree that it should be marked synthetic.

I don't understand what the asm StaticInitMerger does.  From the name I would guess that it
combines all the static_init methods into one.  I would expect this to possibly cause initialization
order problems in some cases, but obviously I'm  not an expert.  In any case if it generates
a new method that should be marked synthetic.  If it removes static_init methods I'd expect
that to cause problems in some circumstances.

I think what is least likely to cause problems would be if aries left the existing static
init methods alone, added a new one for its own purposes, marked synthetic, and modified the
<clinit> method to call the new static init method either before or after all the existing
static init methods, as appropriate.  I'd hope this wouldn't use the StaticInitMerger class
so we wouldn't need to modify or reimplement it.

I'm also not clear whether you want to reimplement the asm serialVersionUID adder so you can
have a clean-room one that marks the field synthetic.  I subclassed it and copied and modified
the relevant method.

> Proxy or weaving code should mark all its changes synthetic
> -----------------------------------------------------------
>
>                 Key: ARIES-669
>                 URL: https://issues.apache.org/jira/browse/ARIES-669
>             Project: Aries
>          Issue Type: Bug
>          Components: Proxy
>    Affects Versions: 0.4
>            Reporter: David Jencks
>
> In order to make it easier for OWB to ignore the additional goo added by the proxy/weaving
stuff, it should all be marked synthetic.  So far I've identified the added SerialVersionUID
field and static_init_* methods as causing problems with the jcdi tck.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message