I wrote: > I would still perform rollbacks for any errors other than the above > expected SQLITE_BUSY cases, of course, since they indicate something > else went wrong (such as running out of disk space). I think it's > safe to say those are all unusual cases though.
Hmm, http://sqlite.org/lang_transaction.html suggests even SQLITE_BUSY might result in automatic transaction rollback. sqlite3_step() suggests the same thing. Other docs suggest that SQLITE_IOERR_BLOCKED is actually returned in such cases (like a writing transaction failing a cache spill attempt). Which is correct? Silently rolling back a transaction automatically is bad enough, but doing that on SQLITE_BUSY would be just plain evil. ----------------------------------------------------------------------------- To unsubscribe, send email to [EMAIL PROTECTED] -----------------------------------------------------------------------------
