Обсуждение: pgsql: Make pg_dumpall's database ACL query independent of hash table o

Make pg_dumpall's database ACL query independent of hash table order.

Previously GRANT order on databases was not well defined, due to the use
of EXCEPT without an ORDER BY.  Add an ORDER BY, adapt test output.

I don't, at the moment, see reason to backpatch this.

Branch
------
master

Details
-------
http://git.postgresql.org/pg/commitdiff/b4fc645787cc7c614c0c97fc9fffacf2bdc6a388

Modified Files
--------------
src/bin/pg_dump/pg_dumpall.c     | 10 ++++++----
src/bin/pg_dump/t/002_pg_dump.pl |  1 +
2 files changed, 7 insertions(+), 4 deletions(-)

On 2016-10-13 01:33:53 +0000, Andres Freund wrote:
> Make pg_dumpall's database ACL query independent of hash table order.
>
> Previously GRANT order on databases was not well defined, due to the use
> of EXCEPT without an ORDER BY.  Add an ORDER BY, adapt test output.
>
> I don't, at the moment, see reason to backpatch this.

Hrmpf, longfin doesn't like this much (as in unstable order).

http://pgbuildfarm.org/cgi-bin/show_log.pl?nm=longfin&dt=2016-10-13%2001%3A39%3A10

Will look into it, probably tomorrow.

Andres

On 10/12/16 10:05 PM, Andres Freund wrote:
> On 2016-10-13 01:33:53 +0000, Andres Freund wrote:
>> Make pg_dumpall's database ACL query independent of hash table order.
>>
>> Previously GRANT order on databases was not well defined, due to the use
>> of EXCEPT without an ORDER BY.  Add an ORDER BY, adapt test output.
>>
>> I don't, at the moment, see reason to backpatch this.
>
> Hrmpf, longfin doesn't like this much (as in unstable order).
>
> http://pgbuildfarm.org/cgi-bin/show_log.pl?nm=longfin&dt=2016-10-13%2001%3A39%3A10
>
> Will look into it, probably tomorrow.

I'm seeing this failure too, locally.

--
Peter Eisentraut              http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

Andres Freund <andres@anarazel.de> writes:
> On 2016-10-13 01:33:53 +0000, Andres Freund wrote:
>> Make pg_dumpall's database ACL query independent of hash table order.

> Hrmpf, longfin doesn't like this much (as in unstable order).

It was failing for me and Peter as well.  Did you generate the new
test in a non-C locale environment?

            regards, tom lane

On 2016-10-13 10:49:14 -0400, Tom Lane wrote:
> Andres Freund <andres@anarazel.de> writes:
> > On 2016-10-13 01:33:53 +0000, Andres Freund wrote:
> >> Make pg_dumpall's database ACL query independent of hash table order.
>
> > Hrmpf, longfin doesn't like this much (as in unstable order).
>
> It was failing for me and Peter as well.  Did you generate the new
> test in a non-C locale environment?

Well, it's not a new test, just an existing one that was dependant on
hash order. But yes, I did check the output on en_US. I guess slapping a
collate C on it would work.  Brr, this test isn't pretty, and hard to
debug.

Andres

On 2016-10-13 09:41:38 -0700, Andres Freund wrote:
> On 2016-10-13 10:49:14 -0400, Tom Lane wrote:
> > Andres Freund <andres@anarazel.de> writes:
> > > On 2016-10-13 01:33:53 +0000, Andres Freund wrote:
> > >> Make pg_dumpall's database ACL query independent of hash table order.
> >
> > > Hrmpf, longfin doesn't like this much (as in unstable order).
> >
> > It was failing for me and Peter as well.  Did you generate the new
> > test in a non-C locale environment?
>
> Well, it's not a new test, just an existing one that was dependant on
> hash order. But yes, I did check the output on en_US. I guess slapping a
> collate C on it would work.  Brr, this test isn't pretty, and hard to
> debug.

Thanks for pushing something to fix this up. Was about to do the same ;)

Andres