public inbox for [email protected]  
help / color / mirror / Atom feed
pgpool: Fix to not call pool_get_transaction_isolation() in elog(DEBUG)
4+ messages / 1 participants
[nested] [flat]

* pgpool: Fix to not call pool_get_transaction_isolation() in elog(DEBUG)
@ 2026-05-19 09:21 Tatsuo Ishii <[email protected]>
  0 siblings, 0 replies; 4+ messages in thread

From: Tatsuo Ishii @ 2026-05-19 09:21 UTC (permalink / raw)
  To: [email protected]

Fix to not call pool_get_transaction_isolation() in elog(DEBUG).

where_to_send_main_replica() and where_to_send_native_replication()
mistakenly called pool_get_transaction_isolation() in
ereport/elog(DEBUG1). Since pool_get_transaction_isolation() issues
do_query() inside, if the code path is executed in an aborted
transaction, do_query() raises a FATAL error. Actually this could
happen when an empty query is executed in an aborted transaction.  (DB
applications/drivers issue an empty query to probe whether connection
to pgpool is active or not).

To fix this, simply delete calls to
pool_get_transaction_isolation(). This of course remove the
information from ereport/elog output, I wasn't able to find
alternative way.

Author: Tatsuo Ishii <[email protected]>

Backpatch-through: v4.5 where where_to_send_main_replica() and
where_to_send_native_replication() are implemented.

Branch
------
V4_5_STABLE

Details
-------
https://git.postgresql.org/gitweb?p=pgpool2.git;a=commitdiff;h=ed230ec5fa7d1832b4fd4abe1d294e20f2d59...

Modified Files
--------------
src/context/pool_query_context.c | 11 ++++-------
1 file changed, 4 insertions(+), 7 deletions(-)



^ permalink  raw  reply  [nested|flat] 4+ messages in thread

* pgpool: Fix to not call pool_get_transaction_isolation() in elog(DEBUG)
@ 2026-05-19 09:21 Tatsuo Ishii <[email protected]>
  0 siblings, 0 replies; 4+ messages in thread

From: Tatsuo Ishii @ 2026-05-19 09:21 UTC (permalink / raw)
  To: [email protected]

Fix to not call pool_get_transaction_isolation() in elog(DEBUG).

where_to_send_main_replica() and where_to_send_native_replication()
mistakenly called pool_get_transaction_isolation() in
ereport/elog(DEBUG1). Since pool_get_transaction_isolation() issues
do_query() inside, if the code path is executed in an aborted
transaction, do_query() raises a FATAL error. Actually this could
happen when an empty query is executed in an aborted transaction.  (DB
applications/drivers issue an empty query to probe whether connection
to pgpool is active or not).

To fix this, simply delete calls to
pool_get_transaction_isolation(). This of course remove the
information from ereport/elog output, I wasn't able to find
alternative way.

Author: Tatsuo Ishii <[email protected]>

Backpatch-through: v4.5 where where_to_send_main_replica() and
where_to_send_native_replication() are implemented.

Branch
------
V4_6_STABLE

Details
-------
https://git.postgresql.org/gitweb?p=pgpool2.git;a=commitdiff;h=87e0336154b521ee31ffeea530e742e0b5214...

Modified Files
--------------
src/context/pool_query_context.c | 11 ++++-------
1 file changed, 4 insertions(+), 7 deletions(-)



^ permalink  raw  reply  [nested|flat] 4+ messages in thread

* pgpool: Fix to not call pool_get_transaction_isolation() in elog(DEBUG)
@ 2026-05-19 09:21 Tatsuo Ishii <[email protected]>
  0 siblings, 0 replies; 4+ messages in thread

From: Tatsuo Ishii @ 2026-05-19 09:21 UTC (permalink / raw)
  To: [email protected]

Fix to not call pool_get_transaction_isolation() in elog(DEBUG).

where_to_send_main_replica() and where_to_send_native_replication()
mistakenly called pool_get_transaction_isolation() in
ereport/elog(DEBUG1). Since pool_get_transaction_isolation() issues
do_query() inside, if the code path is executed in an aborted
transaction, do_query() raises a FATAL error. Actually this could
happen when an empty query is executed in an aborted transaction.  (DB
applications/drivers issue an empty query to probe whether connection
to pgpool is active or not).

To fix this, simply delete calls to
pool_get_transaction_isolation(). This of course remove the
information from ereport/elog output, I wasn't able to find
alternative way.

Author: Tatsuo Ishii <[email protected]>

Backpatch-through: v4.5 where where_to_send_main_replica() and
where_to_send_native_replication() are implemented.

Branch
------
V4_7_STABLE

Details
-------
https://git.postgresql.org/gitweb?p=pgpool2.git;a=commitdiff;h=a483e03152117d22de65f6b5941466643a9de...

Modified Files
--------------
src/context/pool_query_context.c | 11 ++++-------
1 file changed, 4 insertions(+), 7 deletions(-)



^ permalink  raw  reply  [nested|flat] 4+ messages in thread

* pgpool: Fix to not call pool_get_transaction_isolation() in elog(DEBUG)
@ 2026-05-19 09:21 Tatsuo Ishii <[email protected]>
  0 siblings, 0 replies; 4+ messages in thread

From: Tatsuo Ishii @ 2026-05-19 09:21 UTC (permalink / raw)
  To: [email protected]

Fix to not call pool_get_transaction_isolation() in elog(DEBUG).

where_to_send_main_replica() and where_to_send_native_replication()
mistakenly called pool_get_transaction_isolation() in
ereport/elog(DEBUG1). Since pool_get_transaction_isolation() issues
do_query() inside, if the code path is executed in an aborted
transaction, do_query() raises a FATAL error. Actually this could
happen when an empty query is executed in an aborted transaction.  (DB
applications/drivers issue an empty query to probe whether connection
to pgpool is active or not).

To fix this, simply delete calls to
pool_get_transaction_isolation(). This of course remove the
information from ereport/elog output, I wasn't able to find
alternative way.

Author: Tatsuo Ishii <[email protected]>

Backpatch-through: v4.5 where where_to_send_main_replica() and
where_to_send_native_replication() are implemented.

Branch
------
master

Details
-------
https://git.postgresql.org/gitweb?p=pgpool2.git;a=commitdiff;h=b4f38129cfe6b97f7a390a0e056f8b43ed5ad...

Modified Files
--------------
src/context/pool_query_context.c | 11 ++++-------
1 file changed, 4 insertions(+), 7 deletions(-)



^ permalink  raw  reply  [nested|flat] 4+ messages in thread


end of thread, other threads:[~2026-05-19 09:21 UTC | newest]

Thread overview: 4+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2026-05-19 09:21 pgpool: Fix to not call pool_get_transaction_isolation() in elog(DEBUG) Tatsuo Ishii <[email protected]>
2026-05-19 09:21 pgpool: Fix to not call pool_get_transaction_isolation() in elog(DEBUG) Tatsuo Ishii <[email protected]>
2026-05-19 09:21 pgpool: Fix to not call pool_get_transaction_isolation() in elog(DEBUG) Tatsuo Ishii <[email protected]>
2026-05-19 09:21 pgpool: Fix to not call pool_get_transaction_isolation() in elog(DEBUG) Tatsuo Ishii <[email protected]>

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox