Andrew Dunstan wrote:
> Apart from the already known issue of corruption on power loss (which ought
> to be soluble - presumably the commercial heavyweights don't suffer from
> this problem), I can't see any reason to have low expectations of the w32
> port.
Actually, it has been solved by SRA/Tatsuo. Seems Win32 commit()
doesn't actually sync all files during a checkpoint, only the _open_
files. Tatsuo's solution was to remember all files opened since the
last sync() and open/fsync each one of them to simulate sync(). I
assume I will need some of his code, but I am not there yet.
-- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610)
359-1001+ If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square,
Pennsylvania19073