[Rd] File permissions test during R CMD check

Prof Brian D Ripley ripley@stats.ox.ac.uk
Tue, 26 Feb 2002 09:05:13 +0000 (GMT Standard Time)

On Tue, 26 Feb 2002, Kurt Hornik wrote:

> >>>>> Dirk Eddelbuettel writes:
> > I (still) use RCS, and not CVS, for my own projects. R CMD check does
> > not like the 0444 file permissions on files check into RCS.
> > I conjecture that this stems from desire to have the installed files
> > in a state in which they can be edited, but I might be off base ... In
> > case my conjecture holds, would it be possible to relax the
> > restriction during R CMD check, but then install the files as 0644 ?
> Not as long as R CMD build takes what is in the source dirs.
> There are some comments about possible improvements to permission tests
> in the code of check, but in any case all files that are to be installed
> should be at least 644.  We could relax that for e.g. C source or
> headers, but not e.g. for DESCRIPTION.

First comment: there is a --no-perms option to turn the check off.

I don't think there is a problem with DESCRIPTION, since we make the
installed copy with cat and sed.  However, the data file and inst files are
copied and so might have permissions copied too, but the data files have
chmod 644 in (Unix) INSTALL.  The files in inst do need to be checked.
Installing read-only files makes updating packages a lot harder and would
break the current mechanisms for installing updates over existing copies.
But that only argues for the installed copies being at least 644.

There are a number of S users of SCCS/RCS who distribute tarballs with
read-only permissions.  Inevitably I find I need to make some changes and
get obstructed.  That's not supposed to happen with R packages, but it does
and so I do think we want the C source files and headers in an editable

One other minor point: the source tarball will be used on OSes other than
Unix-alikes.  I guess of those only NTFS on Windows NT/2000/XP has a real
file security system (which is based on VMS and is far more elaborate than
0644) and I am not at all clear what would happen if a read-only tarball
was used there, nor how easy it would be to make Rcmd INSTALL cope with
one.  I don't want to begin to think about cross-platform remote-mounted
file systems ....

Brian D. Ripley,                  ripley@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 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch