On Sat, Dec 7, 2013 at 3:02 PM, Robert Haas <robertmhaas@gmail.com> wrote:
>> Leaving it as separate calls sounds good to me as well - but like you I
>> don't like on_user_exit() that much. Not sure if I can up with something
>> really good...
>> on_shmem_exit_phase() or at_shmem_exit_phase() if we go for a function
>> allowing to specify phases, and just before_shmem_exit() for the "user
>> level" things?
>
> Hmm, before_shmem_exit() and on_shmem_exit() doesn't sound bad.
So here's an updated patch that takes that approach. It has a
substantially reduced footprint compared to the previous version, and
probably less chance of breaking third-party code. I also incorporated
your suggestion of renaming on_dsm_detach_cancel to
cancel_on_dsm_detach.
Unless anyone wants to further kibitz the naming here, I'm thinking
this part is ready to commit. I'll rebase and update the remaining
patches after that's done.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company