Re: ICU locale validation / canonicalization

On Tue, 2023-03-14 at 08:08 +0100, Peter Eisentraut wrote:
> Another issue that came to mind:  Right now, you can, say, develop
> SQL
> schemas on a newer ICU version, say, your laptop, and then deploy
> them
> on a server running an older ICU version.  If we have a cutoff beyond
> which we convert ICU locale IDs to language tags, then this won't
> work
> anymore for certain combinations.  And RHEL/CentOS 7 is still pretty
> popular.

If we just uloc_canonicalize() in icu_set_collation_attributes() then
versions 50-53 can support language tags. Patch attached.

One loose end is that we really should support language tags like "und"
in those older versions (54 and earlier). Your commit d72900bded
avoided the problem, but perhaps we should fix it by looking for "und"
and replacing it with "root" while opening, or something.

--
Jeff Davis
PostgreSQL Contributor Team - AWS