Good catch. Yes, needs to by synched (and that could be tricky, with destroy between). Looks to me like destroy destroys the pairs, but does not clear the queues (prior to change) and clear clears _poolList but not _poolMap. Could be I am wrong. I Will look at this some more this eve.
On 6/2/09, Mark Thomas <ma...@apache.org> wrote: > pste...@apache.org wrote: >> Author: psteitz >> Date: Tue Jun 2 02:01:22 2009 >> New Revision: 780905 >> >> URL: http://svn.apache.org/viewvc?rev=780905&view=rev >> Log: >> Ensure that clear() fully clears the pool. > > <snip/> > >> @@ -1293,6 +1293,7 @@ >> } >> } >> destroy(toDestroy); >> + _poolMap.clear(); >> } >> >> /** > > > Still working through the e-mails from over the weekend but on first > inspection that _poolMap.clear() needs to be inside the sync block to > keep it thread safe. Given that, I am having troubling seeing how the > _poolMap isn't already empty at that point. I wonder if we got to the > real root cause of the bug ...? > > I'll get set up to do some testing and report my findings. > > Mark > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org > For additional commands, e-mail: dev-h...@commons.apache.org > > --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org
