R CMD check problem on R 2.6.1 RC

Dear R-developers,

I'm experiencing a problem with having an example run under R 2.6.1 RC
(downloaded and installed today, r43513), which does not occur on R 2.6.0:
The new version of package relaimpo does list package survey under
"Depends", and an example directly (i.e. not only from within a function
of relaimpo) uses function svydesign from package survey. Package survey
is installed and available when working from within the new
R-installation. However, under R CMD check, the example fails with the
message "could not find function svydesign" (backtranslated from German).
It does work fine under all previous versions of R that I tested (2.3.1,
2.5.1, 2.6.0). 

It may be of interest that function boot from package boot (also listed
under Depends, called from within functions of package relaimpo) works
without problems.

Regards, Ulrike

I suppose I should add that this is under Windows (XP).
Regards, Ulrike


Ulrike Gr?mping wrote:

Dear R-developers,

I'm experiencing a problem with having an example run under R 2.6.1 RC
(downloaded and installed today, r43513), which does not occur on R 2.6.0:
The new version of package relaimpo does list package survey under
"Depends", and an example directly (i.e. not only from within a function
of relaimpo) uses function svydesign from package survey. Package survey
is installed and available when working from within the new
R-installation. However, under R CMD check, the example fails with the
message "could not find function svydesign" (backtranslated from German).
It does work fine under all previous versions of R that I tested (2.3.1,
2.5.1, 2.6.0). 

It may be of interest that function boot from package boot (also listed
under Depends, called from within functions of package relaimpo) works
without problems.

On 11/22/2007 10:11 AM, Ulrike Gr?mping wrote:

 > Hi Duncan,
 >
 > yes, I am talking about a new version (currently called 1.99, to be 

made 2.0 on release). I don't think I should attach the file in a note 
to the list - should I ?

 >
 > I've just attached the tar-file for you. Please let me know if I 

should put it somewhere else.

Sending it to me was fine.  I hope you don't mind me taking the 
conversation back to the list.

I don't get an error with 1.99 either, other than this one which isn't 
your fault:

 > * checking DESCRIPTION meta-information ... WARNING
 > Unknown encoding with non-ASCII data
 > Fields with non-ASCII values:
 >   Packaged
 > See the information on DESCRIPTION files in section 'Creating R
 > packages' of the 'Writing R Extensions' manual.

This came because R CMD build put the o-umlaut in your name into the 
Packaged field, i.e. DESCRIPTION has this:

Packaged: Thu Nov 22 16:06:24 2007; Gr?mping

I would call this a bug in our build script, but not one that is serious 
enough to fix before 2.6.1 gets released.

In the check log I see at least one use of svydesign with a warning but 
no error:

 > ### create design with clusters
 > des <- svydesign(id=~subj,data=dat)

Warning in svydesign(id = ~subj, data = dat) :
   No weights or probabilities supplied, assuming equal probability

So I'd come back to guessing that you have survey installed in a 
non-standard place, and R CMD check isn't finding it.

Duncan Murdoch