On Tue, Jan 13, 2026 at 07:01:48PM -0600, Sami Imseih wrote: > Correct.Thanks for spotting this oversight. I think the is_txn_end flag must > be reset in all hook functions that call pgss_store, except in > pgss_ExecutorEnd, > where we just have to check is_txn_end and increment the nesting_level. Also, > I added a PG_TRY/FINALLY inside pgss_ExecutorEnd to ensure we reset the > nesting_level is reset. > > I added a test case for the query you mentioned above.
@@ -1079,35 +1095,47 @@ pgss_ExecutorEnd(QueryDesc *queryDesc)
{
int64 queryId = queryDesc->plannedstmt->queryId;
- if (queryId != INT64CONST(0) && queryDesc->totaltime &&
- pgss_enabled(nesting_level))
+ if (is_txn_end)
+ nesting_level++;
Couldn't it be possible that we reach this stack of execution with
is_txn_end = true but we don't intend it to be so? For example,
imagine first that we reach pgss_ProcessUtility() for a COMMIT
TransactionStmt, second we error without resetting is_txn_end, third a
different portal is executed a with the same backend: we could exit
the executor with is_txn_end set and nesting_level increased but we
did not intend these events to happen.
--
Michael
signature.asc
Description: PGP signature
