To: pgsql-docs@postgresql.org
Subject: Re: Cannot build docs on Ubuntu 10.04? 
In-reply-to: <29895.1331734915@sss.pgh.pa.us> 
References: <COL116-W9647A1E352CAF66550327A3590@phx.gbl>
	<29895.1331734915@sss.pgh.pa.us>
Comments: In-reply-to Tom Lane <tgl@sss.pgh.pa.us>
	message dated "Wed, 14 Mar 2012 10:21:55 -0400"
Date: Thu, 22 Mar 2012 00:12:51 -0400
Message-ID: <17355.1332389571@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

I wrote:
> Maybe the Makefile should just error out if COLLATEINDEX isn't set,
> rather than trying a "fallback" value that configure has already
> determined isn't going to work.

I looked at this a bit closer, and figured out why the Makefile is
trying an apparently pointless case: the test implemented in configure
is wrong!  In at least some installations of docbook, for instance the
ancient one I've got in captivity on my HPUX box, collateindex.pl is
provided in $(DOCBOOKSTYLE)/bin/, but *it is not marked executable*.
This causes AC_PATH_PROGS to not consider it valid.  However, the docs
Makefile is coded to invoke the script via $(PERL) $(COLLATEINDEX),
which is presumably done precisely to be able to cope with
not-executable scripts.

So I think we should adjust the configure test to accept
$(DOCBOOKSTYLE)/bin/collateindex.pl if it is readable, whether or not it
is executable, and only use the Autoconf macro if we fall back to
searching PATH for collateindex.pl.  Having done that, it should be safe
to simplify the Makefile to throw error if COLLATEINDEX isn't set.

			regards, tom lane