Date: Thu, 27 Aug 2026 08:56:34 -0500
From: Nathan Bossart <nathandbossart@gmail.com>
To: pgsql-hackers@postgresql.org
Cc: alvherre@kurilemu.de
Subject: REPACK (ANALYZE) within transaction block segfaults
Message-ID: <apBCEl4hBHlq0axm@nathan>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Archived-At: <https://www.postgresql.org/message-id/apBCEl4hBHlq0axm%40nathan>
Precedence: bulk

I didn't see this reported yet:

    CREATE TABLE t (a INT PRIMARY KEY, b TEXT);
    INSERT INTO t SELECT g, 'v' FROM generate_series(1, 100) g;
    DO $$ BEGIN EXECUTE 'REPACK (ANALYZE) t'; END $$;

This produces the following output:

    WARNING:  transaction left non-empty SPI stack
    HINT:  Check for missing "SPI_finish" calls.
    WARNING:  snapshot 0x8ef018100 still active
    server closed the connection unexpectedly
        This probably means the server terminated abnormally
        before or while processing the request.

Presumably we need to handle transaction blocks a bit like how vacuum()
does.  Or maybe even prevent REPACK (ANALYZE) within a transaction block.

-- 
nathan