[BioC] Error installing (some) Bioconductor packages on Mac OSX under R-3.0.0?

[Dan Tenenbaum]

Hi Peter,


On Wed, Apr 10, 2013 at 4:56 PM, Peter Langfelder
<peter.langfelder at gmail.com> wrote:
> Following up on the topic of failed loading/attaching of Bioconductor
> packages under R-3.0.0 on Mac OS - is this a known bug of the GUI, or
> a problem with the way Bioconductor compiles packages? The user who
> experienced problems loading impute did not have problems loading CRAN
> packages.
>
> Should the developers be notified, or do they already know about the problem?
>

First we need to figure out what's going on. I've posted a query to R-sig-mac
(https://stat.ethz.ch/pipermail/r-sig-mac/2013-April/010058.html)
where someone may be able to help.

The issue seems to involve packages that have some Fortran code.
The binary Mac packages are compiled with a hardcoded path to
/usr/local/lib/libgfortran.2.dylib
which most systems do not have.
However, command-line R seems to look in the path pointed to by
DYLD_FALLBACK_LIBRARY_PATH (which is
/Library/Frameworks/R.framework/Resources/lib) for this file, and
finds it. But the GUI R does not set this variable.

Maybe we can avoid this situation altogether during the package
installation process:

http://bioconductor.org/checkResults/release/bioc-LATEST/impute/petty-buildbin.html

it may be a matter of somehow changing the order of the paths in the
linking command (the line that starts with llvm-gcc).

We're continuiing to look into this and will get back to you when we
have something to report. We'll certainly contact maintainers if there
is anything they need to do.

Thanks,
Dan




> Thanks,
>
> Peter
>
> On Wed, Apr 10, 2013 at 3:29 PM, Dan Tenenbaum <dtenenba at fhcrc.org> wrote:
>> On Wed, Apr 10, 2013 at 3:08 PM, Stephanie M. Gogarten
>> <sdmorris at u.washington.edu> wrote:
>>> Is the error occurring just in the R GUI, or in any R session?  I'm having a
>>> similar problem with DNAcopy when I try to load in the R GUI, but it works
>>> fine in the terminal or in Emacs.
>>>
>>
>> I have the same results; I get the error in the GUI but not in the
>> command line. So I have two workarounds:
>>
>> 1) Use the command line version of R for the time being; and
>> 2) Start the GUI R like this:
>>
>> Open Terminal.app (in /Applications/Utilities)
>>
>> then paste the following:
>> export DYLD_FALLBACK_LIBRARY_PATH=/Library/Frameworks/R.framework/Resources/lib
>> open /Applications/R.app
>>
>> Then loading the impute package should work. I need to get back to you
>> with a better solution.
>>
>> Dan