Re: pg_upgrade fails with non-standard ACL

Greetings,

* Anastasia Lubennikova (a(dot)lubennikova(at)postgrespro(dot)ru) wrote:
> Cool. It seems that everyone agrees on this patch.

Thanks for working on this, I took a quick look over it and I do have
some concerns.

> I attached the updated version. Now it prints a better error message
> and generates an SQL script instead of multiple warnings. The attached test
> script shows that.

Have you tested this with extensions, where the user has changed the
privileges on the extension? I'm concerned that we'll throw out
warnings and tell users to go REVOKE privileges on any case where the
privileges on an extension's object were changed, but that shouldn't be
necessary and we should be excluding those.

Changes to privileges on extensions should be handled just fine using
the existing code, at least for upgrades of PG.

Otherwise, at least imv, the code could use more comments (inside the
functions, not just function-level...) and there's a few wording
improvements that could be made. Again, not a full endorsement, as I
just took a quick look, but it generally seems like a reasonable
approach to go in and the issue with extensions was the only thing that
came to mind as a potential problem.

Thanks,

Stephen