Status: FixProposed Owner: gazzagu...@googlemail.com CC: adrian.l...@gmail.com, marcus.d...@gmail.com Labels: Type-ReportDefect New issue 3986 by gazzagu...@googlemail.com: Bug with Delay during clock rollover http://code.google.com/p/pharo/issues/detail?id=3986 I have identified a bug with Delays that are scheduled close to the millisecond clock rollover that can occur when other delays expire around the same time. The issue appears to be that when scheduling a delay, the resumption time gets set and, before the AccessProtect critical block can be entered, the TimingSemaphore gets signalled, causing the timer process to handle an expired delay. If the clock rolls over before the resumption of the process that was scheduling the original delay, that delay gets scheduled with the resumption time that was determined prior to the clock rollover, hence a delay of around 6 days. A fix is proposed in the attached change set, tested empirically during a rollover with the result of no "rogue" delays - not conclusive but in our production code it appears the problem is resolved. Regards, Gary