On 5-Aug-08, at 3:32 PM, Yonik Seeley wrote:
AutoCommitTest was failing for me a good percentage of the time... the comment suggested that adding another doc after the commit callback would block until the new searcher was registered. But that's not the case. I've hacked the test for now to just sleep(500) after the commit callback.
Fair enough. It is difficult for me to fix this more permenently, since I can't get it to fail on local machines.
I deleted a bunch of email recently so I checked nabble--it seems that in the last month that AutoCommitTest has failed once in Hudson (July 21) and once in the apache build (August 2). That isn't too bad, but I hope that your change eliminates those entirely.
-Mike
