Horton

Re: retire MemoryContextResetAndDeleteChildren backwards compatibility macro

Nathan Bossart <nathandbossart@gmail.com>

View thread View on postgr.es
From: Nathan Bossart <nathandbossart@gmail.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Dagfinn Ilmari Mannsåker <ilmari@ilmari.org>, Alvaro Herrera <alvherre@alvh.no-ip.org>, pgsql-hackers@postgresql.org
Date: 2023-11-14T17:59:17Z
Lists: pgsql-hackers 
On Tue, Nov 14, 2023 at 12:10:41PM -0500, Tom Lane wrote:
> FWIW, I think it's fine to just nuke MemoryContextResetAndDeleteChildren.
> We ask extension authors to deal with much more significant API changes
> than that in every release, and versions where the updated code wouldn't
> work are long gone.  And, as you say, the existence of that separate from
> MemoryContextReset creates confusion, which has nonzero cost in itself.

That is my preference as well.  Alvaro, AFAICT you are the only vote
against removing it completely.  If you feel ѕtrongly about it, I don't
mind going the __attribute__((deprecated)) route, but otherwise, I'd
probably just remove it completely.

-- 
Nathan Bossart
Amazon Web Services: https://aws.amazon.com

Commits

  1. Retire MemoryContextResetAndDeleteChildren() macro.

  2. Simplify code by getting rid of SPI_push, SPI_pop, SPI_restore_connection.

  3. Redefine MemoryContextReset() as deleting, not resetting, child contexts.

  4. tuplestore_donestoring() isn't needed anymore, but provide a no-op