| Plugin Classloading changes between 3.0M3 and 3.0M4? [message #147239] |
Tue, 21 October 2003 10:24  |
Eclipse User |
|
|
|
Originally posted by: adrian_colyer.uk.ibm.com
I'm working on the AJDT plugin project (www.eclipse.org/ajdt). We have an
unusual situation in that the AspectJ compiler is built on top of the JDT
compiler, and so we include our own versions of (self-contained parts of)
org.eclipse.core, org.eclipse.jdt.core, and org.eclipse.jdt.internal
inside a library used by one of our plugins. This library is *not*
exported by the plugin.xml file.
With no changes in any of the AJDT plugins, AspectJ compilation has broken
in 3.0M4 (from 2.0.0 up to 3.0M3 everything was fine). Therefore something
in M4 has caused us to break. Our suspicion is that we are picking up some
(but curiously definitely not all) of the classes from the 3.0M4 versions
of the packages that we have our own private copies of. I also note that
some of the dynamic plug-in registry function arrived in M4...
So, my question is, what changes if any are there in M4 that change class
loading and classpath resolution for plugins? Does anyone have a clue as
to what might be happening?
We've done some initial investigation, but the subtle way in which the
AspectJ compiler fails makes it very difficult to track down the exact
culprit.
Any pointers gratefully received.
Thanks, Adrian.
|
|
|
| Re: Plugin Classloading changes between 3.0M3 and 3.0M4? [message #147958 is a reply to message #147239] |
Wed, 22 October 2003 16:57  |
| Eclipse User |
|
|
|
Originally posted by: John_Arthorne.oti.com_
The short answer is that classloading has not changed between 3.0 M3 and
M4 (as far as I know). Initial dynamic plugin support is available, but
it is off by default (need to specify -dynamic command line arg to turn
it on). None of this has affected classloaders, unless you have
explicitly loaded the Equinox OSGi runtime on top of M4 (a separate
download from the equinox home page). This new runtime uses a different
classloading story with some subtle differences from the old
classloaders. However, if you're just using plain M4 this is not your
problem.
--
Adrian Colyer wrote:
> I'm working on the AJDT plugin project (www.eclipse.org/ajdt). We have an
> unusual situation in that the AspectJ compiler is built on top of the JDT
> compiler, and so we include our own versions of (self-contained parts of)
> org.eclipse.core, org.eclipse.jdt.core, and org.eclipse.jdt.internal
> inside a library used by one of our plugins. This library is *not*
> exported by the plugin.xml file.
>
> With no changes in any of the AJDT plugins, AspectJ compilation has broken
> in 3.0M4 (from 2.0.0 up to 3.0M3 everything was fine). Therefore something
> in M4 has caused us to break. Our suspicion is that we are picking up some
> (but curiously definitely not all) of the classes from the 3.0M4 versions
> of the packages that we have our own private copies of. I also note that
> some of the dynamic plug-in registry function arrived in M4...
>
> So, my question is, what changes if any are there in M4 that change class
> loading and classpath resolution for plugins? Does anyone have a clue as
> to what might be happening?
>
> We've done some initial investigation, but the subtle way in which the
> AspectJ compiler fails makes it very difficult to track down the exact
> culprit.
>
> Any pointers gratefully received.
> Thanks, Adrian.
>
|
|
|
Powered by
FUDForum. Page generated in 0.02357 seconds
Home