On 18/12/06, Nikolay Diakov <[EMAIL PROTECTED]> wrote:
As a solution to communicate variables out of the interpreter, per
thread, we use thread locals that attach java objects to the current
thread. We cache and exchange information among pre and post processors,
as user variables did not help us for some reason (I think they get
I'll look into that - I would have expected them to work...
reinitialized with strings at the beginning of some of the context,
which made them loose our information). Getting data (beanshell
function) looks like:
__BeanShell(((HashMap)((ThreadLocal)(vars.getObject("thread_safe_vars"))).get()).get("intermediate_click_times"))
Perhaps people can use this to alleviate the problem with throwing out
the whole context.
Sounds like a good idea - we could add some standard methods to give
access to thread-local objects. This would simplify the above syntax,
and could be done in Java and/or in the bshinit files.
cheers,
nik
sebb wrote:
> On 18/12/06, Eric Dalquist <[EMAIL PROTECTED]> wrote:
>> sebb,
>>
>> We believe we're running into this bean shell memory leak:
>> http://issues.apache.org/bugzilla/show_bug.cgi?id=40850
>>
>> We have a script we're trying to use to monitor a cluster of servers so
>> it will be running indefinitely. There is a BeanShell listener that runs
>> command line tasks based on server responses. The script runs out of
>> memory rather quickly with the BeanShell listener enabled.
>>
>> I'd like to get your thoughts on the issue. I'm planning on taking the
>> posted source files and making an actual patch out of them to post.
>
> A patch would definitely be better than the full source files as
> currently provided - this should be against branch/rel-2-2.
>
> However, I have some issues with the existing proposals.
>
> The example BeanShellPreProcessor clears the interpreter variable each
> time, so there little point in the init() method - the process()
> method might as well just create the interpreter each time. [But see
> below]
>
> The BeanShellInterpreter cleanup() method should probably not throw
> any Exceptions - it has already logged any problems. But I like the
> idea of having such a method - perhaps it should be called reset()
> instead.
>
> The other major issue is that setting BeanShellInterpreter to null
> will lose any saved context. Seems to me that this is a serious
> restriction.
>
> Not sure if it is possible to prevent the memory leak whilst still
> keeping the interpreter available, but if so, that would be the best
> way to go.
>
> All BeanShell elements could be changed to call cleanup()/reset()
> after each use.
>
> The bshInterpreter itself could be cleared in the BeanShellInterpreter
> class if necessary - as adding it to the other classes makes
> maintenance a bit harder if the leak is later fixed.
>
> To sum up, I think it would be worth trying:
> - add cleanup()/reset() method to BeanShellInterpreter. If necessary
> that can manage resetting the actual interpreter.
>
> - call the cleanup()/reset() method from all the other BeanShell
> classes. Do not clear the BeanShell interpreter instance variable.
>
> S///
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
>
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
