Re: posgres 12 bug (partitioned table)

<div>Hello all!</div><div> </div><div>I just want to point that Npgsql provider for .Net Core builds queries like that (RETURNING xmin) to keep track for concurrency.</div><div>This bug stops us from moving to partitioned tables in Postgres 12 with Npgsql.</div><div> </div><div><a href="https://www.npgsql.org/efcore/index.html">https://www.npgsql.org/efcore/index.html</a></div><div> </div><div>-- <br />С уважением, Павел</div><div> </div><div> </div><div> </div><div>11.08.2020, 18:34, "Tom Lane" &lt;tgl@sss.pgh.pa.us&gt;:</div><blockquote><p>Robert Haas &lt;<a href="mailto:robertmhaas@gmail.com">robertmhaas@gmail.com</a>&gt; writes:</p><blockquote> On Thu, Jun 4, 2020 at 12:57 PM Tom Lane &lt;<a href="mailto:tgl@sss.pgh.pa.us">tgl@sss.pgh.pa.us</a>&gt; wrote:<blockquote> Pavel Biryukov &lt;<a href="mailto:79166341370@yandex.ru">79166341370@yandex.ru</a>&gt; writes:<blockquote> Hello. I'm catch error "virtual tuple table slot does not have system<br /> attributes" when inserting row into partitioned table with RETURNING<br /> xmin</blockquote></blockquote></blockquote><p> </p><blockquote> I wonder whether it's really correct to classify this as a bug. The<br /> subsequent discussion essentially boils down to this: the partitioned<br /> table's children could use any AM, and they might not all use the same<br /> AM. The system columns that are relevant for the heap may therefore be<br /> relevant to all, some, or none of the children. In fact, any fixed<br /> kind of tuple table slot we might choose to use for the parent has<br /> this problem. If all of the children are of the same type -- and today<br /> that would have to be heap -- then using that type of tuple table slot<br /> for the parent as well would make sense. But there's no real reason<br /> why that's the correct answer in general. If the children are all of<br /> some other type, using a heap slot for the parent is wrong; and if<br /> they're all of different types, it's unclear that anything other than<br /> a virtual slot makes any sense.</blockquote><p><br />Well, if we want to allow such scenarios then we need to forbid queries<br />from accessing "system columns" of a partitioned table, much as we do for<br />views. Failing way down inside the executor seems quite unacceptable.<br /><br />                        regards, tom lane</p></blockquote>