Hi David, Thank you for your report.
On Sat, Jul 15, 2017 at 08:58:21AM -0400, David Lee Lambert wrote: > Some time ago, on my main Debian MySQL server, I added an extra > column to the "mysql.user" table of type "timestamp", for auditing > purposes. It didn't cause any trouble at the time, but now when > I try to issue a GRANT statement I get the error > > ERROR 1547 (HY000): Column count of mysql.user is wrong. Expected 42, found > 43. The table is probably corrupted > > Dropping the extra column gets me past the error, but I wish I could > keep that information. In fact, since the error showed up after "security" > updates on the Debian Stable branch, I consider it a REGRESSION. I don't think it's reasonable to consider this a regression. A stable update (whether security or not) by definition must change some behaviour. In general, I think it's reasonable for such updates to assume that users haven't messed with internal data structures. Otherwise, how would a security update that must modify an internal data structure work? In this particular case, it does appear unnecessary that there is an additional check here, but I won't conclude that without knowing exactly why it was done. I'm happy to ask upstream on their view on this, and the reasoning for this change, and see if they think the error can be reduced to a warning in this case. Robie
signature.asc
Description: PGP signature
