Horton

Thread

Commits

  1. Fix xl_heap_lock WAL record field's data type.

  2. Improve concurrency of foreign key locking

  1. infobits_set WAL record struct field is int8

    Peter Geoghegan <pg@bowt.ie> — 2023-04-11T20:13:49Z

    Commit 0ac5ad5134 ("Improve concurrency of foreign key locking") added
infobits_set fields to certain WAL records. However, in the case of
xl_heap_lock, it made the data type int8 rather than uint8.

I believe that this was a minor oversight. Attached patch fixes the issue.

-- 
Peter Geoghegan

  2. Re: infobits_set WAL record struct field is int8

    Andres Freund <andres@anarazel.de> — 2023-04-11T20:48:44Z

    Hi,

On 2023-04-11 13:13:49 -0700, Peter Geoghegan wrote:
> Commit 0ac5ad5134 ("Improve concurrency of foreign key locking") added
> infobits_set fields to certain WAL records. However, in the case of
> xl_heap_lock, it made the data type int8 rather than uint8.
>
> I believe that this was a minor oversight. Attached patch fixes the issue.

Makes sense. Looks like there never was a flag defined for the sign bit,
luckily. I assume you're just going to apply this for HEAD?

Greetings,

Andres Freund

  3. Re: infobits_set WAL record struct field is int8

    Peter Geoghegan <pg@bowt.ie> — 2023-04-11T20:55:50Z

    On Tue, Apr 11, 2023 at 1:48 PM Andres Freund <andres@anarazel.de> wrote:
> Makes sense. Looks like there never was a flag defined for the sign bit,
> luckily. I assume you're just going to apply this for HEAD?

Yes.

I'm also going to rename the TransactionId field to "xmax", for
consistency with nearby very similar records (like
xl_heap_lock_updated).


-- 
Peter Geoghegan