On Thu, Apr 30, 2020 at 12:20 AM Masahiko Sawada
wrote:
> For the part of treating that case as an index corruption I will need
> some time to review because of lacking knowledge of btree indexes. So
> I'll review it later.
I pushed the refactoring patch today. Thanks for the review.
The final
On Wed, Apr 29, 2020 at 11:38 PM Masahiko Sawada
wrote:
> Thank you for the explanation. This makes sense to me.
Pushed both of the fixes.
Thanks!
--
Peter Geoghegan
On Thu, 30 Apr 2020 at 07:29, Peter Geoghegan wrote:
>
> On Tue, Apr 28, 2020 at 12:21 AM Masahiko Sawada
> wrote:
> > For the first fix it seems better to push down the logic to the page
> > deletion code as your 0001 patch does so. The following change changes
> > the page deletion code so
On Wed, 29 Apr 2020 at 01:17, Peter Geoghegan wrote:
>
> On Tue, Apr 28, 2020 at 12:21 AM Masahiko Sawada
> wrote:
> > I agree with both patches.
>
> Thanks for the review.
>
> > For the first fix it seems better to push down the logic to the page
> > deletion code as your 0001 patch does so.
On Tue, Apr 28, 2020 at 12:21 AM Masahiko Sawada
wrote:
> For the first fix it seems better to push down the logic to the page
> deletion code as your 0001 patch does so. The following change changes
> the page deletion code so that it emits LOG message indicating the
> index corruption when a
On Tue, Apr 28, 2020 at 12:21 AM Masahiko Sawada
wrote:
> I agree with both patches.
Thanks for the review.
> For the first fix it seems better to push down the logic to the page
> deletion code as your 0001 patch does so. The following change changes
> the page deletion code so that it emits
On Tue, 28 Apr 2020 at 11:35, Peter Geoghegan wrote:
>
> On Mon, Apr 27, 2020 at 11:02 AM Peter Geoghegan wrote:
> > I would like to backpatch both patches to all branches that have
> > commit 857f9c36cda -- v11, v12, and master. The second issue isn't
> > serious, but it seems worth keeping
On Mon, Apr 27, 2020 at 11:02 AM Peter Geoghegan wrote:
> I would like to backpatch both patches to all branches that have
> commit 857f9c36cda -- v11, v12, and master. The second issue isn't
> serious, but it seems worth keeping v11+ in sync in this area. Note
> that any backpatch theoretically
Attached are two patches, both of which are fixes for bugs in nbtree
VACUUM page deletion.
The first fix for a bug in commit 857f9c36cda. The immediate issue is
that the code that maintains the oldest btpo.xact in the index
accesses the special area of pages without holding a buffer pin. More