From PostgreSQL 18 onwards - TupleDesc struct uses compat_attribute. TupleDescAttr() is the recommended way to access attr now, which is available on postgresql 16

[sanchayanghosh]

Starting from PostgreSQL 18 onwards TupleDesc no longer contains attr. it now uses CompactAttr compact_attr.

In PostgreSQL 16, we still support both attr and compact_attrs.

We need to define a macro that can resolve to both tts_tupleDescriptor->attr[i] and to TupleDescAttr(tts_tupleDescriptor, i)

[sanchayanghosh]

This would also work with releases built for PG17 in #2111

[harudini]

Hi,

I applied your patch but encountered a segmentation fault during the cypher_match regression test.
The backtrace indicates the fault occurs in the function process_update_list, specifically around the call to ExecCloseIndices.

It appears that PostgreSQL 18 no longer tolerates ExecCloseIndices being called twice on the same ResultRelInfo. While in here, ExecCloseIndices is called both inside process_update_list and again within update_entity_tuple, causing a double close that leads to the assertion failure and subsequent segfault.

[sanchayanghosh]

Oh! My bad! Thanks so much for pointing out. Will take a look and fix.

I will run the regression tests again. The last time I ran regression test against master branch and did not see this issue. Will re-run

[jrgemignani]

Oh! My bad! Thanks so much for pointing out. Will take a look and fix.

I will run the regression tests again. The last time I ran regression test against master branch and did not see this issue. Will re-run

Make sure to do a make clean first? Sometimes make gets confused, so I always include clean, just in case.

[sanchayanghosh]

Closing this issue, PGSQL 18 support added in a different PR.