Re: [exim] Build exim from git - xml problem

On 2013-07-18 at 17:28 +0100, John Horne wrote:
> "/usr/share/xml/docbook/stylesheet/nwalsh/1.71.1/xhtml/docbook.xsl"
> compilation error:
> file /home/john/exim/exim-packaging-4.81/release_tree/doc/doc-docbook/MyStyle-txt-html.xsl line 8 element import
> xsl:import : unable to
> load /usr/share/xml/docbook/stylesheet/nwalsh/1.71.1/xhtml/docbook.xsl
> =====================================
>
> I have tried to sort this out, and it seems the file
> 'doc/doc-docbook/OS-Fixups' is where the 'nwalsh' bit is coming from. I
> have changed the file to specify the location of the stylesheets on my
> system (Fedora 17, '/usr/share/sgml/docbook/xsl-stylesheets-1.78.1/').
>
> However, I still get the same error.

Is the error that it can't find the file, or that it's refusing to?

Did you see HowItWorks.txt ? In particular, "XSL INCLUDES"?

> Can anyone provide any insight into this please?

I'm culpable for the OS-Fixups tool and trying to make this more
portable than the original. I'm not an XML person, though.

> Alternatively could a switch be provided to the 'mk_exim_release' script
> so that the docs are not built? It seems that building Exim itself is
> not a problem, but building the docs is and requires other programs as
> well as having static/specific pathnames. I would like a clean Exim
> build, but am happy to run it without the documentation if necessary.

The docs as-are use the portable names and do *not* use fixed paths,
because I fixed that. ;) It might be, if your system has working
catalog systems, that the right fix is to _remove_ the OS fixup for your
OS?

I am extremely reluctant to have a "build a release" option which does
not include spec.txt, because we want to be able to guarantee that
documentation is available for any pseudo-release. If we get away from
that, there will be a cascade of more and more pressure making it harder
to have working docs for the release (making life of the release
engineer harder), and documentation for Exim is a first-class part of
the product, equal in importance to code.

-Phil