All,
On 2004 Oct 14 , at 13.23, Malcolm J. Currie wrote:
> I'm surprised that three mention hlink explicitly. Like Norman, I
> thought hlink was needed for each new hypertext document. There was a
> discussion to simplify (one pass at the end?) to make it more
> efficient,
> but I don't recall it being implemented, nonetheless has muddied the
> waters. From the other perspective, the fact that only three explicit
> dependencies are known may merely indicate that we've not been thorough
> in setting up dependencies, as discussed in other threads this week.
I thought David was suggesting that he'd found only three packages
which used hlink at all, and that none of them declared it explicitly
as a dependency.
He also suggests:
> - kaplibs uses hlink to create an index of the html files, which
> kaplibs then uses to create some html which allows clever searching of
> the kaplibs prologues. This seems a legitimate reason for adding htx
> as a
> build (or should it be sourceset) dependency of kaplibs.
...that only kaplibs has a real need for hlink, since it uses the
output for its own purposes. So kaplibs should declare a dependency on
htx and have a STAR_CHECK_PROGS(hlink) in its configure.ac.
If hlink is required for building the _distributed_ sources -- that is,
for building the set of sources which are compiled into the distributed
sourceset -- then htx should be a sourceset dependency.
Peter notes that:
> Clearly when you're working on a single component hlink-ing can be
> useful
> when you're modifying the documenation, and a pain at other times. The
> moral seems to be that you need both options.
Well yes, sort of. However those few folk who are likely to be
modifying documentation can probably cope wth running hlink themselves,
so I'd be reluctant to add extra complication for just their benefit.
So I vote that we don't bother with hlink in any standard macros,
remove the call to it from ast and (very probably) echomop, and if we
think it necessary just add a final hlink step to the `make world'
rule, possibly via
world: $(ALL_TARGETS)
@HLINK@ ...
How's that?
Norman
--
----------------------------------------------------------------------
Norman Gray : Physics & Astronomy, Glasgow University, UK
http://www.astro.gla.ac.uk/users/norman/ : www.starlink.ac.uk
|