On Fri, Apr 13, 2018 at 9:06 AM, Andres Freund <andres@anarazel.de> wrote:
> On 2018-04-13 12:29:21 +0900, Amit Langote wrote:
>> On 2018/04/13 7:36, Peter Geoghegan wrote:
>> > On Thu, Apr 12, 2018 at 11:47 AM, Peter Geoghegan <pg@bowt.ie> wrote:
>> >> In short, it looks like the tests added to update.sql by commit
>> >> 2f178441 ("Allow UPDATE to move rows between partitions") lead to this
>> >> failure, since I always hit a problem when update.sql is reached. I
>> >> haven't gone to the trouble of digging any deeper than that just yet.
>> >
>> > Without having looked at it in much detail, this seems rather more
>> > likely to be the fault of 2f178441. That was recent enough that it's
>> > easy to believe that I'd be the first to notice it, and actually has
>> > on-disk changes, in the form of ItemPointerSetMovedPartitions().
>>
>> I think you meant f16241bef7cc2 (Raise error when affecting tuple moved
>> into different partition), because it's the one that adds
>> ItemPointerSetMovedPartitions. Commit 2f178441 you quoted added the test
>> that triggered the failure.
>
> Right. Amul, have you looked at this?
>
I have looked into this and found that the issue is in heap_xlog_delete -- we
have missed to set the correct offset number from the target_tid when
XLH_DELETE_IS_PARTITION_MOVE flag is set.
Please find the attached patch does the correction and a big thanks to
Peter for the bug report.
Regards,
Amul