I did some further investigation. While there may be a memory leak,
it could well be a case of it always having been there so that is
worth separate investigation. I did hook yourkit up to it and didn't
see anything in particular that might be a culprit.
However, the reason I'm seeing the problem is this: http://
jira.codehaus.org/browse/MASSEMBLY-194
It's actually a functionality regression when you include unpacked
dependencies in your assembly. A test project is attached that
demonstrates it. It takes about 13s on 2.1, and 1m 30s on 2.2-beta-1
(and is triple the size on output).
So, it looks like the 155 fix can go back in, but I think this
regression needs fixing before release.
Thanks!
Cheers,
Brett
On 31/03/2007, at 11:40 AM, Brett Porter wrote:
> No dice :( I made sure to delete the copy from my local repo so it
> was downloaded again.
>
> I checked the build order, and you were right - the CCE blew it up
> before it got to the problem spot.
>
> I'll try and come up with a test case for you.
>
> - Brett
>
> On 31/03/2007, at 2:32 AM, John Casey wrote:
>
>> Unfortunately, it wasn't that simple. I had actually combined two
>> merges in
>> a single commit to the beta-1 tag...a no-no, I know, but it's
>> fixed now.
>> I've put the newest deployment that excludes MASSEMBLY-155 out
on my
>> people.a.o acct, for your perusal.
>>
>> Let me know what you find out, and thanks for checking this so
>> thoroughly.
>>
>> -john
>>
>> On 3/30/07, Brett Porter <[EMAIL PROTECTED]> wrote:
>>>
>>>
>>> On 31/03/2007, at 12:43 AM, John Casey wrote:
>>>
>>> > I'll try rolling out the MASSEMBLY-155 fix, just to see if that
>>> > makes a
>>> > difference. I can't imagine the ClassCastException fix or the
>>> file-
>>> > mode
>>> > processing chewing up much, though.
>>> >
>>> > Brett: Is it possible that you weren't running out of memory
>>> > previously
>>> > *because* of the CCE?
>>>
>>> I'd need to check the build order, but that's entirely possible -
>>> good point :)
>>>
>>> I'll try it again tomorrow - if you want to point me at a rev# to
>>> rollback I'm happy to do so to avoid you having to muck around
with
>>> deployments.
>>>
>>> Thanks for your patience...
>>>
>>> Cheers,
>>> Brett
>>>
>>>
>>>
--------------------------------------------------------------------
>>> -
>>> 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]
