Re: OK, that's one LOCALE bug report too many... - Mailing list pgsql-hackers

From Tom Lane
Subject Re: OK, that's one LOCALE bug report too many...
Date
Msg-id 26891.975190495@sss.pgh.pa.us
Whole thread Raw
In response to Re: OK, that's one LOCALE bug report too many...  (Peter Eisentraut <peter_e@gmx.net>)
Responses Re: OK, that's one LOCALE bug report too many...  (Bruce Momjian <pgman@candle.pha.pa.us>)
List pgsql-hackers
Peter Eisentraut <peter_e@gmx.net> writes:
> Lamar Owen writes:
>> Ok, let me repeat -- the '--enable-locale' setting will not affect the
>> collation sequence problem on RedHat.  If you set PostgreSQL to use
>> locale, it uses it.  If you configure PostgreSQL to not use locale, the
>> collation set by LANG, LC_ALL, or LC_COLLATE is _STILL_ honored, thanks
>> to the libc used.

> Well, I'm looking at Red Hat 7.0 here and the locale variables are most
> certainly getting ignored in the default compile.  Moreover, at no point
> did strncmp() in glibc behave as you claim.

I'm having a hard time believing Lamar's recollection, also.  I wonder
if there could have been some other factor involved?  One possible line
of thought: a non-locale-enabled compilation, installed to replace a
locale-enabled one, would behave rather inconsistently if run on the
same database used by the locale-enabled version (since indexes will
still be in locale order).  Depending on what tests you did, you might
well think that it was still running locale-enabled.

BTW: as of my commits of an hour ago, the above failure mode is no
longer possible, since a non-locale-enabled Postgres will now refuse to
start up in a database that shows any locale other than 'C' in pg_control.
        regards, tom lane


pgsql-hackers by date:

Previous
From: Peter Eisentraut
Date:
Subject: Re: OK, that's one LOCALE bug report too many...
Next
From: Larry Rosenman
Date:
Subject: tcl/FreeBSD 4.2-STABLE, multiple TCL versions installed