[R-sig-Debian] [Rd] problem with update.packages (PR#13161)

[Dirk Eddelbuettel]

On 14 October 2008 at 07:27, Dirk Eddelbuettel wrote:
| 
| On 14 October 2008 at 12:38, Prof Brian Ripley wrote:
| | The error is on installed.packages(): you have a faulty package installed.
| 
| Confirmed. That explains why my update.r from littler worked as it looked
| only at /usr/local/lib/R/.  Being more explicit about lib.loc I get:
| 
| > update.packages(repos="http://cran.r-project.org", lib.loc="/usr/lib/R/site-library")
| Error in read.dcf(pkgpath, fields = fields) : 
|   Line starting 'unix; ...' is malformed!
| > update.packages(repos="http://cran.r-project.org", lib.loc="/usr/lib/R/library")
| boot :
|  Version 1.2-33 installed in /usr/lib/R/library 
|  Version 1.2-34 available at http://cran.r-project.org
| Update (y/N/c)?  n
| [...]
| > 
| 
| But how do I find out which package this is, and why did R 2.7.2 didn't fall
| over?

Found it -- it was the non-CRAN package 'sg' for the Shogun toolbox.

Soeren: There is a bug the in the DESCRIPTION file of shogun-r version
0.6.3-1: It needs to keep this on one line:

	Built: 2.7.1; i486-pc-linux-gnu; > .Platform[1]
	unix;

or else R's internal updates fail on R 2.8.0 onwards as reported by Simon on
r-sig-debian yesterday.

R-Core: Please close the bug report on R's side, and thanks as always to BDR
for being spot-on in the analysis.

Dirk
 
| Dirk
| 
| | 
| | On Tue, 14 Oct 2008, Dirk Eddelbuettel wrote:
| | 
| | >
| | > On 14 October 2008 at 10:09, Prof Brian Ripley wrote:
| | > | On Tue, 14 Oct 2008, Peter Dalgaard wrote:
| | > |
| | > | > edd at debian.org wrote:
| | > | >> On 14 October 2008 at 09:39, Simon Blomberg wrote:
| | > | >> | Hi, is anyone else getting this error? Could this be a bug?:
| | > | >> | | > update.packages(ask=FALSE)
| | > | >> | Error in read.dcf(pkgpath, fields = fields) : |   Line starting 'unix;
| | > | >> ...' is malformed!
| | > | >> | | > sessionInfo()
| | > | >> | R version 2.8.0 RC (2008-10-12 r46696) | x86_64-pc-linux-gnu
| | > | >> Confirmed in plain R using the same package built this morning from last
| | > | >> night's tarball from Denmark:  Running update.packages() yields:
| | > | >>
| | > | >>> update.packages()
| | > | >> --- Please select a CRAN mirror for use in this session ---
| | > | >> Loading Tcl/Tk interface ... done
| | > | >> Error in read.dcf(pkgpath, fields = fields) :
| | > | >>   Line starting 'unix; ...' is malformed!
| | > | >>
| | > | >>
| | > | >> OTOH I don't have a problem running my 'update.r' script (included in
| | > | >> littler as
| | > | >> /usr/share/doc/littler/examples/update.r) which just downloaded / upgraded
| | > | >> ten packages. It is equivalent to calling
| | > | >>> repos <- "http://cran.r-project.org"
| | > | >>> lib.loc <- "/usr/local/lib/R/site-library"
| | > | >>> update.packages(repos=repos, ask=FALSE, lib.loc=lib.loc)
| | > | >>
| | > | >> and that sequence also works in plain R.  Could someone please dig into
| | > | >> update.packages() ?
| | > | >> Dirk
| | > | >
| | > | > Hmm, some of us are wary of running update.packages() in non-production
| | > | > versions. Is the problem perhaps originating in getCRANmirors() or
| | > | > chooseCRANmirror()?
| | > |
| | > | (Neither of those read DCF files: my expectation is that this is
| | > | available.packages() reading a DCF file downloaded from a CRAN mirror.)
| | > |
| | > | I do use update.packages() daily on pre-2.8.0 and I have not seen a
| | > | problem.  (I keep a separate package installation for test versions.)
| | > |
| | > | Can we please have some useful debugging info, e.g. the result of
| | > | traceback() and which CRAN mirror is being used?  I suspect a problem with
| | > | downloading PACKAGES.gz resulting in a corrupt file (possibly because it
| | > | was corrupt on the mirror).
| | >
| | > Quite possible. This is a fresh failure picking the 'US (MI)' mirror:
| | >
| | >> update.packages()
| | > --- Please select a CRAN mirror for use in this session ---
| | > Loading Tcl/Tk interface ... done
| | > Error in read.dcf(pkgpath, fields = fields) :
| | >  Line starting 'unix; ...' is malformed!
| | >> traceback()
| | > 6: read.dcf(pkgpath, fields = fields)
| | > 5: .readPkgDesc(lib, fields)
| | > 4: installed.packages(lib.loc = lib.loc)
| | > 3: NROW(instPkgs)
| | > 2: old.packages(lib.loc = lib.loc, contriburl = contriburl, method = method,
| | >       available = available, checkBuilt = checkBuilt)
| | > 1: update.packages()
| | >>
| | >
| | > But if I then re-run chooseCRANmirror() and pick Austria, it still fails:
| | >
| | >> chooseCRANmirror()
| | >> update.packages()
| | > Error in read.dcf(pkgpath, fields = fields) :
| | >  Line starting 'unix; ...' is malformed!
| | >>
| | >
| | > yielding the same traceback.
| | >
| | > Dirk
| | >
| | > -- 
| | > Three out of two people have difficulties with fractions.
| | >
| | 
| | -- 
| | 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
| 
| -- 
| Three out of two people have difficulties with fractions.

-- 
Three out of two people have difficulties with fractions.