Horton
Menu

Why is citext/regress failing on hamerkop?

Thomas Munro <thomas.munro@gmail.com>

View thread
From: Thomas Munro <thomas.munro@gmail.com>
To: pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2024-05-11T01:14:46Z
Lists: pgsql-hackers 
For example, 'i'::citext = 'İ'::citext fails to be true.

It must now be using UTF-8 (unlike, say, Drongo) and non-C ctype,
given that the test isn't skipped.  This isn't the first time that
we've noticed that Windows doesn't seem to know about İ→i (see [1]),
but I don't think anyone has explained exactly why, yet.  It could be
that it just doesn't know about that in any locale, or that it is
locale-dependent and would only do that for Turkish, the same reason
we skip the test for ICU, or ...

Either way, it seems like we'll need to skip that test on Windows if
we want hamerkop to be green.  That can probably be cribbed from
collate.windows.win1252.sql into contrib/citext/sql/citext_utf8.sql's
prelude... I just don't know how to explain it in the comment 'cause I
don't know why.

One new development in Windows-land is that the system now does
actually support UTF-8 in the runtime libraries[2].  You can set it at
a system level, or for an application at build time, or by adding
".UTF-8" to a locale name when opening the locale (apparently much
more like Unix systems, but I don't know what exactly it does).  I
wonder why we see this change now... did hamerkop have that ACP=UTF-8
setting applied on purpose, or if computers in Japan started doing
that by default instead of using Shift-JIS, or if it only started
picking UTF-8 around the time of the Meson change somehow, or the
initdb-template change.  It's a little hard to tell from the logs.

[1] https://www.postgresql.org/message-id/CAC%2BAXB10p%2BmnJ6wrAEm6jb51%2B8%3DBfYzD%3Dw6ftHRbMjMuSFN3kQ%40mail.gmail.com
[2] https://learn.microsoft.com/en-us/windows/apps/design/globalizing/use-utf8-code-page

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Small refactoring around ExecCreateTableAs().

  2. Skip citext_utf8 test on Windows.

  3. Fix lost Windows socket EOF events.

  4. Enable routine running of citext's UTF8-specific test cases.

  5. Drops in the CreateProcess calls for Win32 (essentially wrapping up the