On Wed, Dec 10, 2025 at 01:04:31AM +0200, Heikki Linnakangas wrote:
> Yes please. I'm also happy to help if you'd like. I'm eagerly waiting to see
> if the "Widen MultiXactOffset to 64 bits" commit caused any new
> cross-version pg_ugprade test failures.

Okay, I'll see about doing something.

By the way, there is also a consequence in using the first test name
in the REGRESS list for the database name: we cannot use the same name
for the "setup" phase of all these test modules, or the gin, typcache
and nbtree modules would conflict with each other because they would
use the same database name.

By the way, it seems to me that even under meson the two nbtree tests
do not compete with each other: they are executed in a serialized
fashion.  Parallel mode can be enforced with a schedule, but it's just
to say that a DROP EXTENSION in each SQL file should work as well.

>> By the way, REGRESS in modules/nbtree/Makefile is missing
>> nbtree_half_dead_pages..  meson.build has both tests.
> 
> That one is on me..

I am not going to touch this line, but I don't mind doing something
about it as I am touching the same area.  Another thing: the nbtree
module is missing a .gitignore.
--
Michael

Attachment: signature.asc
Description: PGP signature

Reply via email to