Re: [aspectwerkz-dev] tracing to the console
"Alex" <alex-he/e8N7AIe/[email protected]>
| Newsgroups | gmane.comp.java.aspectwerkz.devel |
|---|---|
| Message-ID | <[email protected]> |
Sure, this one should work fine to intercept user hacked ClassLoader subclasses. Alex ----- Original Message ----- From: "Eugene Kuleshov" <eu-E/[email protected]> To: <[email protected]> Sent: Wednesday, July 14, 2004 12:19 PM Subject: Re: [aspectwerkz-dev] tracing to the console > Alex, Jonas, > > That all is godd to know but it is all AW internals. What I'm looking for > is to bind and advice to the following pointcut > > execution(* java.lang.ClassLoader+.new(..)) > > It will be ok to do not have advice at the very top class loader. > > regards, > Eugene > > > > Exact > > There is indeed some ways to alter the java.lang.ClassLoader as well with > > JVMTI, but then it makes things complex to support offline deployments. > > Alex > > > > > ----- Original Message ----- > > From: "Jonas Boner" <[email protected]> > > To: <[email protected]>; <[email protected]> > > Sent: Tuesday, July 13, 2004 10:13 PM > > Subject: RE: [aspectwerkz-dev] tracing to the console > > > > > > > >>The CL is modified in a pre-step if you run HotSwap mode. This is > > > > something we want to get away from since it is slow (requires -Xdebug) and > > overly complex (and differs in impl between the JVMs). > > > >>The preferred choice is the JMAPI in jrockit and JVMTI in 1.5. > >> > >>I don't think that I want to add a generic join point shadow to the > > > > defineClass in the CL since it will add additional overhead that would not > > be much of use. This could of course be user configurable. So, sure, it can > > be added as an option if you are running the HotSwap mode (or the offline). > > > >>But as I said, we are moving away from this option. > >> > >>/Jonas > >> > >>-----Original Message----- > >>From: eu [mailto:eu-E/[email protected]] > >>Sent: Tue 7/13/2004 8:48 PM > >>To: [email protected] > >>Cc: > >>Subject: Re: [aspectwerkz-dev] tracing to the console > >> > >>Jonas Boner wrote: > >> > >> > >>>Then you would have to do post-compilation of rt.jar. > >> > >> Hmm. That is confusing. > >> > >> I see that ClassLoader class is being instrumented (at least when I > >>start application using aspectwerkz.bat), why not leave joinpoint in > > > > there? > > > >> regards, > >> Eugene > >> > >> > >> > >>>-----Original Message----- > >>>From: eu [mailto:eu-E/[email protected]] > >>>Sent: Tue 7/13/2004 7:51 PM > >>>To: [email protected] > >>>Cc: > >>>Subject: Re: [aspectwerkz-dev] tracing to the console > >>> > >>>Alex wrote: > >>> > >>> > >>> > >>>>The AOP container is designed to avoid to hook directly in the > > > > classloader > > > >>>>class. I don't want to hook in classloaders. It was already complex to > >>>>enable class load time weaving. > >>> > >>> > >>> Hmm. Is there are any way to put a custom advice to something like > >>>execution(* java.lang.ClassLoader+.new(..)) ? > >