Re: Non-reproducible AIO failure

[Konstantin Knizhnik]

One more update: with the proposed patch (memory barrier before 
`ConditionVariableBroadcast` in `pgaio_io_process_completion` and 
replacing bit fields with `uint8`) the problem is not reproduced at my 
system during 50000 seconds.