Обсуждение: Re: [COMMITTERS] pgsql: Permit dump/reload of not-too-large >1GB tuples

Поиск
Список
Период
Сортировка

Re: [COMMITTERS] pgsql: Permit dump/reload of not-too-large >1GB tuples

От
Tom Lane
Дата:
Alvaro Herrera <alvherre@alvh.no-ip.org> writes:
> Permit dump/reload of not-too-large >1GB tuples

I apologize for not having paid close enough attention earlier, but:
this patch is absolutely unacceptable for the back branches and MUST
be reverted there.  Adding another field to StringInfoData is an ABI
change that will certainly break large numbers of extensions.  We
can't expect those to get recompiled for minor releases.
        regards, tom lane



Re: [COMMITTERS] pgsql: Permit dump/reload of not-too-large >1GB tuples

От
Alvaro Herrera
Дата:
Tom Lane wrote:
> Alvaro Herrera <alvherre@alvh.no-ip.org> writes:
> > Permit dump/reload of not-too-large >1GB tuples
> 
> I apologize for not having paid close enough attention earlier, but:
> this patch is absolutely unacceptable for the back branches and MUST
> be reverted there.  Adding another field to StringInfoData is an ABI
> change that will certainly break large numbers of extensions.  We
> can't expect those to get recompiled for minor releases.

Oh, I see the problem now -- it can (and frequently is) stack allocated,
not just palloc'ed using the StringInfo api, and changing the size
breaks that.  Rats.  I'll revert tomorrow.

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



Re: [HACKERS] Re: [COMMITTERS] pgsql: Permit dump/reload ofnot-too-large >1GB tuples

От
Alvaro Herrera
Дата:
Alvaro Herrera wrote:
> Tom Lane wrote:
> > Alvaro Herrera <alvherre@alvh.no-ip.org> writes:
> > > Permit dump/reload of not-too-large >1GB tuples
> > 
> > I apologize for not having paid close enough attention earlier, but:
> > this patch is absolutely unacceptable for the back branches and MUST
> > be reverted there.  Adding another field to StringInfoData is an ABI
> > change that will certainly break large numbers of extensions.  We
> > can't expect those to get recompiled for minor releases.
> 
> Oh, I see the problem now -- it can (and frequently is) stack allocated,
> not just palloc'ed using the StringInfo api, and changing the size
> breaks that.  Rats.  I'll revert tomorrow.

Done.

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