Re: Crash report for some ICU-52 (debian8) COLLATE and work_mem values

Peter Eisentraut <peter(dot)eisentraut(at)2ndquadrant(dot)com> writes:
> On 8/5/17 23:26, Tom Lane wrote:
>> I'm quite disturbed though that the set of installed collations on these
>> two test cases seem to be entirely different both from each other and from
>> what you reported. The base collations look generally similar, but the
>> "keyword variant" versions are not comparable at all.

> The differences you are seeing are probably because we intentionally
> removed the keyword variants for ICU 4.2:
> <https://www.postgresql.org/message-id/E1ddzFn-0005x5-N0%40gemulon.postgresql.org>.

After looking more closely, I think I'd mistaken keyword variants for
something else. As an example, I get these "af-*" collations installed
by initdb when using ICU 52.1:

af-x-icu
af-u-co-standard-x-icu
af-u-co-eor-x-icu
af-u-co-search-x-icu
af-NA-x-icu
af-NA-u-co-standard-x-icu
af-NA-u-co-eor-x-icu
af-NA-u-co-search-x-icu
af-ZA-x-icu
af-ZA-u-co-standard-x-icu
af-ZA-u-co-eor-x-icu
af-ZA-u-co-search-x-icu

while with ICU 57.1, the set is

af-x-icu
af-u-co-standard-x-icu
af-u-co-emoji-x-icu
af-u-co-eor-x-icu
af-u-co-search-x-icu

So they added emojis (I'm with Peter G that we could do without installing
that by default) ... but what became of the af-NA and af-ZA collations?
If I were a user who'd adopted one of those as a database collation,
I'd be seriously unhappy to have them go away in a later PG release.

I think we'd be well advised to filter the set of installed-by-default
collations rather strongly, in hope of avoiding such problems. For
starters, do we really need the keyword variants at all? People who
know what those are for can create their own collations, and take
their own risks of the feature disappearing in later ICU releases.

regards, tom lane