R help - Mar 2004 - Different missing links on Windows in 'check' vs. 'install'

Hello ...

Using R-1.9.0 alpha, I'm having some problem getting a few packages to
pass check under Windows - specifically with the 'missing link(s)'
section
of the package install phase.

I started trying to track down how the missing link was showing up as in
some cases I could not see why the link was considered missing.  For
instance, in one package it was looking for 'makeViewers' although when
the packages is loaded in R, 'makeViewers' is even in the search path
(by
way of a dependent package which was loaded initially).

In trying to figure out how these things are determiend, I noticed that in
Windows there was a difference in the reported missing links when one does
'Rcmd check' vs 'Rcmd install' (or 'Rcmd install
--build').  In the
example above, using either of the 'install' methods results in that
link
not being reported as missing (and if I put in an intentionally missing
link it gets picked up as such), but 'check' reports it as missing. 
I'm
wondering what the difference in environment is between check & install on
Windows as that might help me to figure out why check reports these as
missing links.

Thanks
-Jeff

On Mon, 8 Mar 2004 18:20:29 -0500 (EST), you wrote:
>In trying to figure out how these things are determiend, I noticed that in
>Windows there was a difference in the reported missing links when one does
>'Rcmd check' vs 'Rcmd install' (or 'Rcmd install
--build').  In the
>example above, using either of the 'install' methods results in that
link
>not being reported as missing (and if I put in an intentionally missing
>link it gets picked up as such), but 'check' reports it as missing. 
I'm
>wondering what the difference in environment is between check & install
on
>Windows as that might help me to figure out why check reports these as
>missing links.
This is something we should fix. I haven't checked the code, but I
believe the rule is that check will only find links to base packages,
but install will find links to any package installed on the system.

You can avoid the errors in check by using the
\link[package:topic]{foo} syntax, to tell R where to find the link:
but you need to give the filename of the Rd file as the topic, not an
alias to it.

Duncan Murdoch

Note that on Windows only packages in the same library are checked for 
links, as on Windows the HTML help cannot link across libraries.  (This is 
done under Unix via symbolic links.)

By default check installs in a private library, so you may well see 
(correct) reports of missing links under Windows.  You can check an 
already installed version, though.

On Mon, 8 Mar 2004, Jeff Gentry wrote:
> Hello ...
> 
> Using R-1.9.0 alpha, I'm having some problem getting a few packages to
> pass check under Windows - specifically with the 'missing link(s)'
section
> of the package install phase.
> 
> I started trying to track down how the missing link was showing up as in
> some cases I could not see why the link was considered missing.  For
> instance, in one package it was looking for 'makeViewers' although
when
> the packages is loaded in R, 'makeViewers' is even in the search
path (by
> way of a dependent package which was loaded initially).
Not relevant, as this is Perl code and no R process is running.
> In trying to figure out how these things are determiend, I noticed that in
> Windows there was a difference in the reported missing links when one does
> 'Rcmd check' vs 'Rcmd install' (or 'Rcmd install
--build').  In the
> example above, using either of the 'install' methods results in
that link
> not being reported as missing (and if I put in an intentionally missing
> link it gets picked up as such), but 'check' reports it as missing.
I'm
> wondering what the difference in environment is between check & install
on
> Windows as that might help me to figure out why check reports these as
> missing links.
-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595