Horton

Re: pg_dump / copy bugs with "big lines" ?

Alvaro Herrera <alvherre@2ndquadrant.com>

View thread View on postgr.es
From: Alvaro Herrera <alvherre@2ndquadrant.com>
To: Daniel Verite <daniel@manitou-mail.org>
Cc: Tomas Vondra <tomas.vondra@2ndquadrant.com>, Craig Ringer <craig@2ndquadrant.com>, Tom Lane <tgl@sss.pgh.pa.us>, Robert Haas <robertmhaas@gmail.com>, Jim Nasby <Jim.Nasby@bluetreble.com>, Ronan Dunklau <ronan.dunklau@dalibo.com>, pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2016-11-28T17:28:40Z
Lists: pgsql-hackers 
I just wrote:

> The big advantage of your v3 patch is that it can be backpatched without
> fear of breaking ABI, so I've struggled to maintain that property in my
> changes.  We can further tweak in HEAD-only; for example change the API
> to use Size instead of int.  I think that would be desirable, but let's
> not do it until we have backpatched this one.

One thing I just noticed while trying to backpatch this is that we can
do so only to 9.5, because older branches do not have
MemoryContextAllocExtended().  They do have MemoryContextAllocHuge(),
but the caller in heaptuple.c wants zeroed memory too, so we'd need to
memset; I think that could get us back to 9.4.

9.3 and older is not possible because we don't have "huge alloc" there.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

Commits

  1. Fix overflow check in StringInfo; add missing casts

  2. Permit dump/reload of not-too-large >1GB tuples