Fix order of operations in ExecEvalFieldStoreDeForm().
authorTom Lane <[email protected]>
Thu, 29 Jun 2023 14:19:10 +0000 (10:19 -0400)
committerTom Lane <[email protected]>
Thu, 29 Jun 2023 14:19:10 +0000 (10:19 -0400)
commitd0ab203bc192382053d3782062bef6872bdc0e53
treeb04c43d1eb1c023da74313bc346c1cbf5f3f24c4
parentf42844069755c0a6761fede783b8acb063e880a5
Fix order of operations in ExecEvalFieldStoreDeForm().

If the given composite datum is toasted out-of-line,
DatumGetHeapTupleHeader will perform database accesses to detoast it.
That can invalidate the result of get_cached_rowtype, as documented
(perhaps not plainly enough) in that function's API spec; which leads
to strange errors or crashes when we try to use the TupleDesc to read
the tuple.  In short then, trying to update a field of a composite
column could fail intermittently if the overall column value is wide
enough to require toasting.

We can fix the bug at no cost by just changing the order of
operations, since we don't need the TupleDesc until after detoasting.
(Other callers of get_cached_rowtype appear to get this right already,
so there's only one bug.)

Note that the added regression test case reveals this bug reliably
only with debug_discard_caches/CLOBBER_CACHE_ALWAYS.

Per bug #17994 from Alexander Lakhin.  Sadly, this patch does not fix
the missing-values issue revealed in the bug discussion; we'll need
some more work to cover that.

Discussion: https://postgr.es/m/17994-5c7100b51b4790e9@postgresql.org
src/backend/executor/execExprInterp.c
src/test/regress/expected/rowtypes.out
src/test/regress/sql/rowtypes.sql