[Springframework-developer] RE: About CGLIB2

[Chris N. <ch...@si...>]

jürgen höller wrote:
> Hibernate 2.1 final with CGLIB2 underneath has just been released. This
> poses the question whether we should still release our 1.0 M4 with CGLIB1
> support. Of course, our current codebase works nicely with Hibernate 2.1
> as long as you don't use CGLIB-based AOP proxies (i.e. proxy full target
> objects instead of just interfaces), so we don't *have* to update
> immediately.
[snip]
> CGLIB2 is at RC1 in the meantime, so building on it now would be viable.
> In the face of this situation, does everybody still prefer deferred CGLIB2
> support at the time of our RC1? I vote for migrating already in M4, to
> allow for CGLIB-based AOP proxies in combination with Hibernate 2.1
> promptly. After all, typical Spring users will choose to work with the
> most current recommended Hibernate version.
> 
> Not that this was a really important criterion, but nevertheless: Chris,
> when do you expect CGLIB2 final?

Wow, that's a big thread! I'll try to answer everything in this mail.

Here's the current status:

Hibernate 2.1 Final ships with CGLIB 2.0 RC1.

CGLIB 2.0 Final will probably be released within a week. At the same
time, CGLIB *1.1* will also be released. This will be exactly the same
as CGLIB 1.0 except it will use ASM 1.4 instead of 1.3, which will allow
it to be used simutaneously with CGLIB 2.0 (no classname are shared
between CGLIB 1.0 and 2.0 for this reason).

With CGLIB 1.1 available, your decision to use CGLIB 2.0 for AOP can
be independant of whether you want to support Hibernate 2.1, I think.


Rod Johnson wrote:
> Chris, I guess I'd better take another look at your examples of how we
> could do away with CGLIB "optimization" in favour of CGLIB2 lazy
> interceptors or whatever they're called. Is there any documentation on
> the changes in CGLIB and how to take best advantage of them, besides
> your excellent emails here :-)

No "article"-type documentation yet. I'll try to answer whatever
questions you have in the meantime.


Colin Sampaleanu wrote:
> Yes, getting rid of that bloated beast will clear up 500k!
> BCELAttributes and BCELAttributeWriter from the Sandbox can still be
> accessed from CVS history when we want to do the new versions...

It should be pretty easy to port these things over to CGLIB/ASM. ASM 1.4
also has some support for the JSR 175 classfile attributes, in case you
want to be future-compatible.


jürgen höller wrote:
> On a related note, the iBATIS Database Layer 1.2.8 (as used by our
> JPetStore version) requires CGLIB 1.0. But as I said, this doesn't
> matter as long as you don't use CGLIB-based AOP proxies (which we
> don't in our JPetStore version). It's worth mentioning in the release
> notes, though.

I think the release of CGLIB 1.1 will make this moot. 


Rod Johnson wrote:
> Anyone know if Clinton has plans to upgrade to CGLIB 2?

Yes, I've been in contact with Clinton and version 2 of iBATIS should be
using CGLIB 2.

Let me know if you have any more questions.

Chris