R-2.5.0 install

I am receiving the following errors when trying to install
gcrma and a number of other Bioconductor packages when
the installWithVers flag is set to be TRUE.

 >source("http://www.bioconductor.org/getBioC.R");
 >getBioC("gcrma", installWithVers=T);

Hi, Anna.  You will probably want to follow the instructions here:

http://bioconductor.org/docs/install-howto.html

Sean

Sean Davis <sdavis2 at mail.nih.gov> writes:

Anna Lobley wrote:

I am receiving the following errors when trying to install
gcrma and a number of other Bioconductor packages when
the installWithVers flag is set to be TRUE.

 >source("http://www.bioconductor.org/getBioC.R");
 >getBioC("gcrma", installWithVers=T);

Hi, Anna.  You will probably want to follow the instructions here:

http://bioconductor.org/docs/install-howto.html

Actually, getBioC and biocLite are fairly straight forward wrappers to
install.packages.  They pass '...' arguments to install.packages and
so Anna's use of installWithVers is valid and expected to work.

And for me it does.

Anna, can you try the following:

      rep = "http://bioconductor.org/packages/2.0/bioc"
      install.packages("Biobase", repos=rep, installWithVers=TRUE)

I would also be curious to understand your motivation for wanting
installWithVers.  BioC packages are developed for a particular R
release and in general I would not expect BioC packages developed
older R's to work with current R's.

+ seth

Seth Falcon | Computational Biology | Fred Hutchinson Cancer Research Center
http://bioconductor.org

Update:

Seth Falcon <sfalcon at fhcrc.org> writes:

And for me it does.

Almost.  With a clean package library, I end up with an error
installing matchprobes because library.dynam can't find a shared
library called matchprobes_1.8.0.  Since other packages that have
native code are installing with version ok, I suspect this is a
fixable bug in matchprobes.

Anna, if you still are having problems, we need more detail about what
packages are already installed, loaded, etc.  It might make sense to
try this with fresh library... on Linux this might look like:

      mkdir TMPLIB1
      export R_LIBS=`pwd`/TMPLIB1
      R
      source("http://biocondcutor.org/biocLite.R")
      biocLite("gcrma", installWithVers=TRUE)

But depending on your setup, you may want to instect the output of
.libPaths() to see if you are picking up other libraries.

+ seth

Seth Falcon | Computational Biology | Fred Hutchinson Cancer Research Center
http://bioconductor.org

Update:

Seth Falcon <sfalcon at fhcrc.org> writes:

And for me it does.

Almost.  With a clean package library, I end up with an error
installing matchprobes because library.dynam can't find a shared
library called matchprobes_1.8.0.  Since other packages that have
native code are installing with version ok, I suspect this is a
fixable bug in matchprobes.

Indeed it is.  The package has

.First.lib <- function(lib, pkgname, where) {
   ## load the compiled code
   library.dynam(pkgname, pkgname, lib)

and that should be

   library.dynam("matchprobes", pkgname, lib)

(That's not an uncommon mistake, BTW: EBarrays, puma and twilight of BioC 
seem to share it.)

Also, .First.lib will only ever tbe called with two args, and 'where' is 
unused.

BTW, I have BioC-1.9 and BioC-2.0 in separate library trees rather than 
use versioned installs: although the idea and code for the latter came 
from the BioC group, they are little used and have some rough edges.

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

Prof Brian Ripley <ripley at stats.ox.ac.uk> writes:

Indeed it is.  The package has

.First.lib <- function(lib, pkgname, where) {
  ## load the compiled code
  library.dynam(pkgname, pkgname, lib)

and that should be

  library.dynam("matchprobes", pkgname, lib)

Thanks!  I've fixed this for the release and devel versions of
matchprobes.

(That's not an uncommon mistake, BTW: EBarrays, puma and twilight of
BioC seem to share it.)

Indeed.  I will contact the maintainers of those packages.

+ seth

Seth Falcon | Computational Biology | Fred Hutchinson Cancer Research Center
http://bioconductor.org