On 13.06.2011 21:31, Tom Lane wrote:
> So I finally started actually reading the SSI changes, and I am a tad
> distressed by this:
>
> diff --git a/src/include/access/twophase_rmgr.h b/src/include/access/twophase_rmgr.h
> index a541d0f..1c7d8bb 100644
> --- a/src/include/access/twophase_rmgr.h
> +++ b/src/include/access/twophase_rmgr.h
> @@ -23,8 +23,9 @@ typedef uint8 TwoPhaseRmgrId;
> */
> #define TWOPHASE_RM_END_ID 0
> #define TWOPHASE_RM_LOCK_ID 1
> -#define TWOPHASE_RM_PGSTAT_ID 2
> -#define TWOPHASE_RM_MULTIXACT_ID 3
> +#define TWOPHASE_RM_PREDICATELOCK_ID 2
> +#define TWOPHASE_RM_PGSTAT_ID 3
> +#define TWOPHASE_RM_MULTIXACT_ID 4
> #define TWOPHASE_RM_MAX_ID TWOPHASE_RM_MULTIXACT_ID
>
> extern const TwoPhaseCallback twophase_recover_callbacks[];
>
> What was the rationale for changing the assignments of existing 2PC IDs?
As far as I can tell it was for purely cosmetic reasons, to have lock
and predicate lock lines together.
> So far as I can tell, that breaks pg_upgrade (if there are any open
> prepared transactions) for no redeeming social benefit.
Surely pg_upgrade can't work anyway if there's any open prepared
transactions in the database. We're not going to guarantee to keep all
the data structures we write in two-phase state files unchanged over
major releases. If pg_upgrade is not checking for prepared transcations
at the moment, such a check should probably should be added.
-- Heikki Linnakangas EnterpriseDB http://www.enterprisedb.com