[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
[aspectj-dev] RE: BCEL 5.1

I agree with Wes that we shouldn't do this for 1.1final.  
I don't see myself having the time to carefully 
swap out BCEL versions under a release candidate -- especially
because I'd be direly tempted to do BCEL cleanup at the same
time (we did do some special handling of the getSuperClass() issue).  

If it can wait, I believe that the right time to 
upgrade to 5.1 will be just before tackling enhancement 
36069, "Correct source information while inlining", which
will require grafting a JSR45 layer onto BCEL.  


-----Original Message-----
From: Wes Isberg [mailto:wes@xxxxxxxxxxxxxx] 
Sent: Tuesday, 20 May, 2003 1:00 pm
To: aspectj-dev@xxxxxxxxxxx
Subject: BCEL 5.1

BCEL 5.1 was released April 25 and appears to be
both source- and binary-compatible with the current tree,
with preliminary tests showing no change in compiler
behavior.  There seems to have been no Apache announcement.

The bug list for bcel is


The most relevant of the bugs fixed is that
JavaClass.getSuperClass() will return Object
rather than null when Object is the immediate superclass
(but null when getting a superclass of Object). I'm
not sure if any of our tests check this, but I seem
to recall Erik working around BCEL's actual behavior
in this case.


Other fixes pertain to classes we don't use, but I can
imagine a weaving class loader in the same namespace
wanting to use them:
- an API conflict between the verifier and JDK 1.4 Throwable
- NPE in the classloader

There are other possibly-destablizing bugs outstanding:

Many of these appear nicely investigated, to possibly
occur in the old version as well, and to have reasonable
patches that we could incorporate.

I'm assuming we should upgrade after the 1.1 final
release, but would do the upgrade beforehand if Jim
or Erik wanted and no committers objected.