Re: Revert: Remove useless self-joins *and* -DREALLOCATE_BITMAPSETS make server crash, regress test fail.

David Rowley <dgrowleyml@gmail.com>

From: David Rowley <dgrowleyml@gmail.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Andrew Dunstan <andrew@dunslane.net>, Richard Guo <guofenglinux@gmail.com>, jian he <jian.universality@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2024-05-07T23:03:16Z
Lists: pgsql-hackers
On Wed, 8 May 2024 at 10:55, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> David Rowley <dgrowleyml@gmail.com> writes:
> > REALLOCATE_BITMAPSETS was invented for this and IMO, it found exactly
> > the problem it was invented to find.
>
> Not in a way that gives me any confidence that we found *all* the
> problems.

Here are some statements I believe to be true:
1. If REALLOCATE_BITMAPSETS is defined then modifications to a
Bitmapset will make a copy and free the original.
2. If a query runs successfully without REALLOCATE_BITMAPSETS and
Assert fails due to an invalid Bitmapset when REALLOCATE_BITMAPSETS is
defined, then we have > 1 pointer pointing to the same set and not all
of them are being updated when the members are added/removed.

Given the above, I can't see what Bitmapset sharing problems we won't
find with REALLOCATE_BITMAPSETS.

Can you share the exact scenario you're worried that we won't find so
I can understand your concern?

David



Commits

  1. Make left-join removal safe under -DREALLOCATE_BITMAPSETS.

  2. Revert: Remove useless self-joins

  3. Remove mention of nchar