Re: is allow_nonpic_in_shlib still useful?

From: Noah Misch <noah(at)leadboat(dot)com>
To: Peter Eisentraut <peter_e(at)gmx(dot)net>
Cc: pgsql-hackers(at)postgresql(dot)org
Subject: Re: is allow_nonpic_in_shlib still useful?
Date: 2012-12-15 13:22:11
Message-ID: 20121215132211.GB10608@tornado.leadboat.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Sat, Dec 15, 2012 at 01:23:38AM -0500, Peter Eisentraut wrote:
> In the plperl and plpython makefiles we look for a shared library of
> libperl or libpython, and if it's not found, we check for
> allow_nonpic_in_shlib, and if that's yes, then we proceed anyway.
> Apparently, and IIRC, this was set up in a time when those shared
> libraries were not available through standard builds, but I think that
> hasn't been the case for quite a while.
>
> The only platforms where we set allow_nonpick_in_shlib is linux and
> freebsd/i386 (presumably an obsolescent combination). Are there any
> Linux builds that don't supply the required shared libraries?

I can't recall such a system. On x86_64, GNU ld would reject the resulting
text relocations anyway.

> I suspend this hack isn't useful anymore and ought to be removed.

Agreed. On !allow_nonpic_in_shlib systems, the effect appears to be that we
quietly skip the plperl build, despite --with-perl, when the Perl build lacks
a shlib. This seems unhelpful; I think --with-perl should result in either an
error or a built plperl. Likewise for plpython. The coincidentally-better
build behavior under allow_nonpic_in_shlib should become unconditional.

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Erik Rijkers 2012-12-15 13:24:05 Re: small pg_basebackup display bug
Previous Message Magnus Hagander 2012-12-15 13:10:02 Re: small pg_basebackup display bug