Re: Re: Missing rows with index scan when collation is not "C" (PostgreSQL 9.5)

On Wed, Mar 23, 2016 at 12:13 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> I wrote:
>> I extended my test program to be able to check locales using ISO-8859-x
>> encodings.  RHEL6 shows me failures in a set of locales that is remarkably
>> unlike the set it fails on for UTF8 (though good ol de_DE manages to fail
>> in both encodings, as do a few others).  I'm not sure what that implies
>> for the underlying bug(s).
>
> Closer analysis says that all of the cases where only utf8 is reported to
> fail are in fact because there is no iso8859 equivalent locale on my
> machine.  Many of the cases where only iso8859 is reported to fail are
> just chance passes due to not having randomly generated a failure case;
> you can reduce the odds of that by passing strcolltest a repeat count
> larger than 1.  There remain, however, a few locales in which it seems
> that indeed iso8859 is broken and utf8 is not; ru_RU being the most
> prominent example.
>
> In short, the problem is actually worse in non-UTF8 locales.

I guess that's not terribly surprising.  If the glibc maintainers
haven't managed to get this right for UTF-8 locales, I can't imagine
why they would have been more careful for non-UTF-8 locales that - I
would guess - get less use.

Are you still in information-gathering more, or are you going to issue
a recommendation on how we should proceed here, or what?

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company