[mythtv] cvs.sql

[Andy Davidoff]

You don't seem to've addressed this in your other messages on the subject,
so I'm replying here for context...

What about "ALTER TABLE CHANGE"?

MODIFY won't work if it's referencing a column that doesn't exist, so this
suggests that a CHANGE needs to occur before the MODIFY, but if the CHANGE
occurs before the column is ADDed, it will incorrectly fail.

If I understand the (perceived and AFAWK never exhibited) problem correctly,
you're afraid that two updates made to CVS will not be applied identically
on hosts who have zero or one updates already applied.

I wrote an applet to do this using the kind of logic we don't have in MySQL.
It's a quick hack and can be resubmitted if necessary.  Do we really want to
keep hacking at cvs.sql?  Even if we don't use my applet, I think we should
come up with a simpler (safer) method for CVS SQL updates, or at least a
better way to indicate failure to the user.


#if Chris Palmer /* Mar 30, 16:13 */
> So, I'm trying to reorganize cvs.sql so that it can run straight thru
> without errors on anyone's system until it gets to something that it
> doesn't need to do, such as trying to add a new field that already exists.
#endif /* mythtv at zencow.com */