Re: ICU for global collation

Hi,

On Fri, Jan 21, 2022 at 10:44:24AM +0100, Peter Eisentraut wrote:
>
> Here is a second preparation patch: Change collate and ctype fields to type
> text.
>
> I think this is useful by itself because it allows longer locale names. ICU
> locale names with several options appended can be longer than 63 bytes.
> This case is probably broken right now. Also, it saves space in the typical
> case, since most locale names are not anywhere near 63 bytes.

I totally agree.

> From 1c46bf3138ad42074971aa3130142236de7e63f7 Mon Sep 17 00:00:00 2001
> From: Peter Eisentraut <peter(at)eisentraut(dot)org>
> Date: Fri, 21 Jan 2022 10:01:25 +0100
> Subject: [PATCH] Change collate and ctype fields to type text

+ collversionstr = TextDatumGetCString(datum);
+
actual_versionstr = get_collation_actual_version(collform->collprovider, collcollate);
if (!actual_versionstr)
{
@@ -1606,7 +1616,6 @@ pg_newlocale_from_collation(Oid collid)
(errmsg("collation \"%s\" has no actual version, but a version was specified",
NameStr(collform->collname))));
}
- collversionstr = TextDatumGetCString(collversion);

Is that change intended? There isn't any usage of the collversionstr before
the possible error when actual_versionstr is missing.

Apart from that the patch looks good to me, all tests pass and no issue with
building the doc either.