RE: Bug in logical decoding with DDL and subtransactions - Mailing list pgsql-hackers

From Hayato Kuroda (Fujitsu)
Subject RE: Bug in logical decoding with DDL and subtransactions
Date
Msg-id OS7PR01MB18317AFD6A46C4B4A7C4A461EF5962@OS7PR01MB18317.jpnprd01.prod.outlook.com
Whole thread
In response to RE: Bug in logical decoding with DDL and subtransactions  (Bingshuai Li <lucian1412@outlook.com>)
Responses Re: Bug in logical decoding with DDL and subtransactions
List pgsql-hackers
Hi,

Thanks for updating the patch.
My idea was to call the cleanup function only once. A cleanup function can accept
a list of sub transactions, and checks them while iterating the tuplecids of the
top transaction. Per my understanding we can use the binary search because the list
of sub transactions are sorted by the logical XID. Attached .txt implements the idea.

> For cfbot, reattaching v5 for master only.  The REL_* attachments in my
> previous message are alternative backpatch versions, not a patch
> series.  Cfbot applied the master patch successfully, then tried to
> apply the REL_14_STABLE version on top of it and reported conflicts.

TIPS: attaching .txt is a known workaround [1].


[1]: https://wiki.postgresql.org/wiki/Cfbot

Best regards,
Hayato Kuroda
FUJITSU LIMITED


Attachment

pgsql-hackers by date:

Previous
From: Alexandre Felipe
Date:
Subject: Re: WAL segment file descriptor leak on read errors can PANIC the server
Next
From: Heikki Linnakangas
Date:
Subject: Re: [PATCH] Fix vacuum_delay_point happening inside lock