Re: [HACKERS] HP Unix 11.00 Compiler error. - Mailing list pgsql-general

From Tom Lane
Subject Re: [HACKERS] HP Unix 11.00 Compiler error.
Date
Msg-id 1885.990731873@sss.pgh.pa.us
Whole thread Raw
In response to HP Unix 11.00 Compiler error.  (Chris Bowlby <excalibur@hub.org>)
Responses Re: [HACKERS] HP Unix 11.00 Compiler error.  (Chris Bowlby <excalibur@hub.org>)
List pgsql-general
Chris Bowlby <excalibur@hub.org> writes:
> The configure script errors out looking
> for _eprintf(), which apparently is a now deprecated GCC
> attempt at a compatibility layer. This tells me I really
> can't use the binary build made by the HP-UX porting archive
> folks.

FWIW, I've seen eprintf link failures on HPUX 10.20 as well.  I think
the key to avoiding it is that you have to build all the components with
the same compiler (all gcc, or all HP's).  If you want to link against a
precompiled ssl library then that will determine your choice of compiler.

> Cannot compile PostgreSQL and C++ library (--with-CXX) with GCC.
> Can't find <string> class in STL. Suspect packaging
> job done by HP-UX porting archive: it moves GCC from
> /usr/local to /opt, and a few things didn't make the
> trip smoothly.

I agree.  I don't have any trouble building the C++ library here,
using either hand-built gcc or HP's cc.  I let gcc install itself
in the usual place, ie, /usr/local.

The Porting Archive guys do good work, but their stuff gives headaches
if you want to mix and match it with stuff you build yourself.  They're
way too eager to mess around with install locations.

I'd recommend rebuilding gcc in a vanilla configuration, using the
porting archive gcc just for bootstrap purposes.

            regards, tom lane

pgsql-general by date:

Previous
From: Tom Lane
Date:
Subject: Re: Trouble with strange OUTER JOIN syntax
Next
From: Chris Bowlby
Date:
Subject: Re: [HACKERS] HP Unix 11.00 Compiler error.