Обсуждение: pg_dump with disable trigger

Hello All,

I think I found bug on pg_dump 9.6.8.

The bug arise when using pg_dump with --disable-trigger option.
The command just :
pg_dump -f /tmp/test_disable_trigger.sql -a -b --disable-triggers -t some_table some_database

On the resulted script, it simply did

ALTER TABLE public.some_table DISABLE TRIGGER ALL;

COPY command

ALTER TABLE public.some_table ENABLE TRIGGER ALL;

This become a problem, when the destination table already has Disabled user triggers, it gets enabled by the dump script.

It also altered another kind of trigger like Triggers firing on replica only, get altered to ordinary triggers.

On Tue, Apr 10, 2018 at 4:17 PM, Soni M <diptatapa@gmail.com> wrote:

Hello All,

I think I found bug on pg_dump 9.6.8.

The bug arise when using pg_dump with --disable-trigger option.
The command just :
pg_dump -f /tmp/test_disable_trigger.sql -a -b --disable-triggers -t some_table some_database

On the resulted script, it simply did

ALTER TABLE public.some_table DISABLE TRIGGER ALL;

COPY command

ALTER TABLE public.some_table ENABLE TRIGGER ALL;

This become a problem, when the destination table already has Disabled user triggers, it gets enabled by the dump script.

I would classify this as a limitation of the concept as opposed to a bug.​  I was going to say a bit more but I'm not that fluent with disabled triggers mode in dump/restore setups.  I would advise using pg_restore and leaving the disable triggers option of the pg_dump command.  It lacks sufficient information to write an appropriate script for an unknown target server.

If you have issues with pg_restore those would be more easily be deemed as bugs.

FWIW we have a "SET LOCAL" command that results in GUC variables being restored to their state post-transaction; we'd need a similar command for pg_dump to work properly.

It also altered another kind of trigger like Triggers firing on replica only, get altered to ordinary triggers.

​This seems like a bug - though I'm not going to be confirming the behavior and you haven't provided evidence.

​

David J.