Andreas Hauser <andreas-postgresql(at)creative-memory(dot)de> writes:
> david.rowley wrote @ Wed, 23 Sep 2015 10:57:25 +1200:
>> This seems related to this
>> http://www.postgresql.org/message-id/9A28C8860F777E439AA12E8AEA7694F8010F6F3F@BPXM15GP.gisp.nec.co.jp
>>
>> work_mem, I believe, must be higher than 1GB to trigger this.
> I can confirm that it works when work_mem is reduced to 1GB.
OK, that seems to confirm the diagnosis. Thanks.
That patch thread seems to have gone off into the weeds, but I'll make
sure something gets done about it before 9.5beta1.
regards, tom lane