strcoll and strxfrm do not always agree in -CURRENT

[Thomas Munro]

On 24 March 2016 at 11:36, Thomas Munro <munro at ip9.org> wrote:
> Hi,
>
> The PostgreSQL hackers wrote a test program to study a collation bug
> on another operating system.  The test generates random strings and
> checks if strxfrm and strcoll agree on their ordering, according to
> each collation available on the system.  I decided to try it out on
> FreeBSD and found that it passes on 10.2, but fails miserably on
> 11.0-CURRENT.
>
> The test program is here:
>
> http://www.postgresql.org/message-id/19132.1458688784@sss.pgh.pa.us
>
> I had to increase MAXXFRMLEN to (MAXSTRLEN * 10) for it to work.

Oops, I missed this updated version which doesn't require that
modification.  The result is the same: pass on 10.2, fail on 11.0.

http://www.postgresql.org/message-id/31913.1458747836@sss.pgh.pa.us

Best regards,

Thomas Munro