Re: postgres 7.2 make check problem

From: Michael Robinton <michael(at)bizsystems(dot)com>
To: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
Cc: <pgsql-general(at)postgresql(dot)org>
Subject: Re: postgres 7.2 make check problem
Date: 2002-03-10 17:27:48
Message-ID: Pine.LNX.4.33.0203100918370.29853-100000@pandora.is.bizsystems.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

On Sun, 10 Mar 2002, Tom Lane wrote:

> "Michael" <michael(at)insulin-pumpers(dot)org> writes:
> > I am upgrading from 7.1 -> 7.2
> > ./configure --enable-multibyte=UNICODE
>
> Was your 7.1 installation built with multibyte support?
> (I'm betting not.)
>
That's right. The old version is not mulitbyte and the new one needs to be
for newer applications that require UNICODE.

> > make check >
> /test/regress/./tmp_check/install//usr/local/pgsql/bin/pg_encoding: >
> undefined sy mbol: pg_char_to_encoding initdb: pg_encoding failed
>
> I believe what is happening here is that the dynamic loader is trying
> to link to the installed version of libpq.so (viz, your 7.1 version,
> which I'm guessing has no multibyte support) instead of the newly
> built version that's been placed in the temporary installation.
> pg_regress tries to override the dynamic loader's search path by setting
> LD_LIBRARY_PATH, but that does not work on all platforms. In
> particular, on Linux I believe the RPATH information placed into the
> executables takes precedence over LD_LIBRARY_PATH. So, because you've
> got already-installed libraries sitting at the same place where the
> 7.2 libraries will ultimately be installed, those get picked up in
> preference to the right ones.
>
> This is pretty much a catch-22 situation :-(. AFAICS the only way we
> could make pre-installation "make check" proof against this problem on
> Linux is to not set RPATH in the executables; which would certainly make
> the finished installation less reliable, so it seems like a bad idea
> to do it just to make "make check" work.
>
> If you've got an idea how to get around this problem we'd love to hear
> about it. In the meantime I think you will have to install at least the
> multibyte-aware libpq.so before you can do "make check". Or, if you are
> feeling too paranoid to do that, you could rebuild 7.2 with a --prefix
> pointing to someplace different than where 7.1 is installed, so that the
> RPATH search doesn't find the 7.1 library.
>

Well..... my solution was to do

make install
make check

I've already done data dump, I can always go back and reinstall the older
version.

That seems to have worked. I still must check that the applications all
work ok.

Michael

In response to

Browse pgsql-general by date

  From Date Subject
Next Message Francisco Reyes 2002-03-10 18:10:21 Wal_buffers memory utilization
Previous Message Tom Lane 2002-03-10 16:22:16 Re: Postgres not starting at boot(FreeBSD) - startup script not releasing