On Fri, Sep 30, 2022 at 04:15:24PM -0400, Tom Lane wrote: > In view of the recent mess around bigint relfilenodes, it seems to me > that we shouldn't move forward with widening AclMode unless somebody > runs down which structs will get wider (or more aligned) and how much > that'll cost us. Maybe it's not a problem, but it could do with an > explicit look at the point.
The main one I see is AclItem, which increases from 12 bytes to 16 bytes. AFAICT all of the catalogs that store aclitem arrays have the aclitem[] column marked extended, so they are compressed or moved out-of-line as needed, too. The only other structs I've spotted that make use of AclMode are InternalGrant and InternalDefaultACL. I haven't identified anything that leads me to believe there are alignment problems or anything else comparable to the issues listed in the relfilenode thread [0], but I could be missing something. Did you have something else in mind you think ought to be checked? I'm not sure my brief analysis here suffices. [0] https://postgr.es/m/CA%2BTgmoaa9Yc9O-FP4vS_xTKf8Wgy8TzHpjnjN56_ShKE%3DjrP-Q%40mail.gmail.com -- Nathan Bossart Amazon Web Services: https://aws.amazon.com