On 03/27/2016 12:43 AM, Christophe Pettus wrote:
> On Mar 26, 2016, at 7:40 AM, Andrew Dunstan <andrew@dunslane.net> wrote:
>> It would be nice if we could find a less broad brush approach to dealing with the issue.
> I don't know how doable this is, but could we use the existing mechanism of marking an index invalid if it contains
anenum type to which a value was added, and the transaction was rolled back? For the 90% use case, that would be
acceptable,I would expect.
>
The more I think about this the more I bump up against the fact that
almost anything we do might want to do to ameliorate the situation is
going to be rolled back. The only approach I can think of that doesn't
suffer from this is to abort if an insert/update will affect an index on
a modified enum. i.e. we prevent the possible corruption from happening
in the first place, as we do now, but in a much more fine grained way.
cheers
andrew