Re: Move defaults toward ICU in 16?

On 08.03.23 06:55, Jeff Davis wrote:
> On Fri, 2023-03-03 at 21:45 -0800, Jeff Davis wrote:
>>>>     0002: update template0 in new cluster (as described above)
> 
> I think 0002 is about ready and I plan to commit it soon unless someone
> has more comments.

0002 seems fine to me.

> I'm holding off on 0001 for now, because you objected. But I still
> think 0001 is a good idea so I'd like to hear more before I withdraw
> it.

Let's come back to that after dealing with the other two.

>>>>     0003: default initdb to use ICU
> 
> This is also about ready, and I plan to commit this soon after 0002.

This seems mostly ok to me.  I have a few small comments.

+        default, ICU obtains the ICU locale from the ICU default collator.

This seems to be a fancy way of saying, the default ICU locale will be 
set to something that approximates what you have set your operating 
system to.  Which is what we want, I think.  Can we say this in a more 
user-friendly way?

+static void
+check_icu_locale()

should be check_icu_locale(void)

+       if (U_ICU_VERSION_MAJOR_NUM >= 54)
+       {

If we're going to add more of these mysterious version checks, could we 
add a comment about what they are for?

However, I suspect what this chunk is doing is some sort of 
canonicalization/language-tag conversion, which per the other thread, I 
have some questions about.

How about for this patch, we skip this part and just do the else branch

+           icu_locale = pg_strdup(default_locale);

and then put the canonicalization business into the canonicalization 
patch set?