Howard Chu wrote:
Howard Chu wrote:
I'm seeing this failure with both BDB 4.2.52 and BDB 4.6.21, and no failure in RE23, so it must be something changed between RE23 and RE24, but so far looking at the diffs I don't see anything that could have caused this change in behavior.
Looking back at 2.4.5 and 2.4.6, I got the same failure in 2.4.6. That means it's not a new regression in 2.4.7 at least. I'm still testing 2.4.5 (using the tests in 2.4.6) to see if it fails. At this point, since it's not new in 2.4.7, we should track it in the ITS. I may just hack up a quick workaround for this until I can isolate the real problem.
Using the 2.4.7 tests, with 3 more searches added to the mix, I got the same failure in 2.4.5 and RE23. So it seems this is really a much older bug. Not sure what to do about it at the moment. Again, we've been relying on BerkeleyDB's lock subsystem to prevent this sort of event from ever happening, and clearly that's not working. I'll send a query to the BerkeleyDB folks about it.