[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[equinox-dev] Re: Question for Equinox class loading experts
- From: Hal Hildebrand <hal.hildebrand@xxxxxxxxxx>
- Date: Thu, 28 Jun 2007 11:21:49 -0700
- Delivered-to: email@example.com
- Thread-index: Ace5q39Svb2FRCWeEdyu7AAbY5LlQwABbHtT
- Thread-topic: Question for Equinox class loading experts
- User-agent: Microsoft-Entourage/11.3.3.061214
More from Andy:
I have tried using PackageAdmin.refreshPackages() to force unlinking and
the Bundle to go away but that seems to be a fairly destructive operation
which causes havoc for us. What I'm not sure of is whether someone is going
to tell me the behaviour is expected unless you call refreshPackages(). I
can see that it might be, although that's not very useful :(
On 6/28/07 10:41 AM, "Hal Hildebrand" <hal.hildebrand@xxxxxxxxxx> wrote:
> Forwarding this message from Andy Piper, from the Spring/OSGi mailing list.
> We are seeing an issue in equinox whereby if we install a bundle,
> uninstall it and install it again the new bundle does _not_ get a new
> classloader if the bundle is leaking memory. I think it works ok if
> the old bundle's classes are all de-referenced, but in a complex
> system that's quite a tall order.
> We would expect the installation of a new bundle to _always_ result
> in a new classloader being created, the fact that it isn't seems like
> a bug to us.
> Note that if we use a different name when we install the bundle again
> we _do_ get a new classloader.
> Any comments from the equinox experts appreciated!