Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Subject: Htmlhelp virtually impossible to build?
Date: Tue, 4 Oct 2005 12:00:02 +0100
Message-ID: 
 <E7F85A1B5FF8D44C8A1AF6885BC9A0E4CC3143@ratbert.vale-housing.co.uk>
Thread-Topic: Htmlhelp virtually impossible to build?
Thread-Index: AcXI0sUzDFB580NdTxOpvOJ0zko3Wg==
From: "Dave Page" <dpage@vale-housing.co.uk>
To: <pgsql-hacker@postgresql.org>
Cc: <pgsql-docs@postgresql.org>

In order to build the docs for the Windows distribution of PostgreSQL
and all platforms supported bt pgAdmin, I need to run 'make htmlhelp'
from doc/src/sgml, which generates html doc pages and htmlhelp indexes,
project file and toc.

Unfortunately this generally seems to fail on the xsltproc step when it
comes to building the toc etc - on 32 bit machines it always seems to
bail out with an invalid pointer error in glibc, whilst on 64 bit boxes,
prior to 8.0.4 it generally worked OK, given 10 minutes or so to
complete. With 8.0.4 however, I get a segfault on 2 different fc4 boxes
I've tried. On 32bit Solaris, I get complaints (followed by a core dump)
along the lines of 'No "en" localization of "foo" exists (which I guess
might just be my environment). For 8.0.4, I finally managed to run the
xsltproc step successfully on Windows XP (using the same version of
xsltproc as on the fc4 boxes).

Anyhoo, I realise this is most likely largely the fault of xsltproc, but
is there anything we can do to improve matters ourselves - for example
by breaking up the processing into chapters rather than doing the whole
manual at once?

Regards, Dave