D. Richard Hipp wrote: > On Jul 23, 2008, at 1:08 PM, Brad House wrote: > >> I'm just investigating an issue now. This is the first ever >> incident of a corrupt database we've had on a few thousand >> installations, > > Have you read the background information at > > http://www.sqlite.org/atomiccommit.html > > See especially section 9.0: Things That Can Go Wrong.
Yes, I've read that. We're using the VFS layer that sqlite provides for windows (since XP-E is just XP with a bunch of dlls and auxiliary applications removed). The database always resides on the same disk as the software accessing it (no network transfers). The database file wouldn't have been manipulated outside of our application. According to your section 9, that leaves a 'rouge' process (read: virus), or buffers not actually being flushed to disk (either because of a disk controller issue or a FlushFileBuffers() issue). Obviously other options exist such as hardware failure (RAM, harddrive), or an SQLite bug. Hardware failure definitely hasn't been ruled out here. The main reason for reporting this issue is to make sure if there is an issue, there is enough 'history' of it to justify researching it. I'm definitely not blaming SQLite at this point. Thanks. -Brad _______________________________________________ sqlite-users mailing list [email protected] http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
