List:Falcon Storage Engine« Previous MessageNext Message »
From:Christopher Powers Date:November 10 2008 3:28pm
Subject:Re: Review fix for Bug#39696, serial log corruption
View as plain text  
Kevin Lewis wrote:
>>> Shouldn't this comment read;
>>>
>>> -    // Flush record data, if any, and force the creation of a new 
>>> serial log window
>>> +    // Flush record data, if any, and force the creation of a new 
>>> serial log record
>>
>> Actually, both are true, but your suggestion is clearer.
> 
> A SerialLogWindow is always 1Mb.  I do not see how starting a new 
> SerialLogRecord will start a new window. The outer loop in 
> SRLUpdateRecords::append just starts a new SerialLogRecord, not a new 
> window.
>     
>   for (RecordVersion *record = records; record;)
>     {
>     START_RECORD(srlUpdateRecords, "SRLUpdateRecords::append");

True, but SerialLog::flush() is called with forceNewWindow=true at the 
bottom of the outer loop, which results in a call to 
SerialLog::createNewWindow().

Not sure why we do this, actually.

Thread
Review fix for Bug#39696, serial log corruptionChristopher Powers9 Nov
  • Re: Review fix for Bug#39696, serial log corruptionKevin Lewis9 Nov
    • Re: Review fix for Bug#39696, serial log corruptionChristopher Powers9 Nov
      • Re: Review fix for Bug#39696, serial log corruptionKevin Lewis10 Nov
        • Re: Review fix for Bug#39696, serial log corruptionChristopher Powers10 Nov