David,
On 2004 Oct 14 , at 10.51, David Berry wrote:
>> Hmm, htx is listed as depending on star2html, but that's only because
>> the htx configure.ac declares STAR_LATEX_DOCUMENTATION(sun188), which
>> includes that dependency automatically. If that line were to be
>> removed, then SUN/188 wouldn't be installed by default (big deal), but
>> we would then be free to declare a (fake) dependency of star2html on
>> htx, which would be enough to cause htx to be built and installed
>> first. How does that sound?
>
> Should we not be up-front about dependencies, and state dependencies on
> htx where they exist rather than relying on a hidden away back-door
> approach?
Yes, if there's only a few, and (more importantly) they're not implicit
in a macro (as the almost universal dependency on star2html is implicit
in STAR_LATEX_DOCUMENTATION).
I confess I'm a bit uncertain about where hlink comes in to the grand
scheme of things. I vaguely thought that it was always (supposed to
be) run after installing .htx docs, and that was why it was (rather,
ought to be) a dependency on star2html. But as I say, running hlink
after every .htx installation doesn't seem entirely sensible, so I'm
happy if it has nothing to do with star2html at all!
Anyone in proud possession of a Clue here (ie, not me) is quite welcome
to edit STAR_LATEX_DOCUMENTATION if they believe it's doing the wrong
or inadequate thing.
> Having just grepped for hlink, the only components which depend on htx
> seem to be ast, kaplibs and echomop. Would it not be simpler just to
> add
> a build dependency on htx to each of these three components?
Is this possibly because the the original authors or adaptors of these
packages had a monopoly on htx-Clue?
Norman
--
----------------------------------------------------------------------
Norman Gray : Physics & Astronomy, Glasgow University, UK
http://www.astro.gla.ac.uk/users/norman/ : www.starlink.ac.uk
|