[Rd] R-3.1.0 OSX Snow Leopard installs old binary

Simon Urbanek simon.urbanek at r-project.org
Fri Apr 18 04:08:49 CEST 2014


On Apr 17, 2014, at 9:52 PM, Gábor Csárdi <csardi.gabor at gmail.com> wrote:

> On Thu, Apr 17, 2014 at 9:43 PM, Simon Urbanek <simon.urbanek at r-project.org> wrote:
> [...]
> The Suggests failure has nothing to do with BioC. Only packages listed in Depends/Imports are required for a package to work so there is no guarantee for any packages in Suggests to be available - hence the package should not break if they are not available - that's the whole point of Suggests. If you list it in Depends/Imports then it won't even get to the check if those packages are not available - it won't build at all. I didn't look at the dependencies in this particular case, but one reason to use Suggests is to break dependency loops: if A depends on B and B on A, then there is no way to install them, so typically A suggests B and B depends on A so that A can be installed and checked first without B and then B checked with A and finally A with B.
> 
> I would naively think that if you install A and B _together_, then they should be fine. At least this is how dependencies work on various Linux distributions, AFAIK.

They cannot be installed together, R doesn't have a concept of "together" since it doesn't separate copying and parse/eval stages so you cannot "pre-install" the packages and then run them through R to create the binary. Therefore dependencies are always sequential.


>  If A breaks without B then it makes such bootstrapping impossible - we found some packages with this issue, that's why mentioned this - I don't know if that's the case with igraph or not.
> 
> As for BioC, the builds for BioC are independent of CRAN, so CRAN doesn't build BioC packages and thus their availability is subject to manual intervention - on the OS X build machine there is currently no automated way to track BioC packages, but we're working on it.
> 
> So this effectively means that if I Import/Depend/Suggest etc. a BioC package in igraph, then igraph will likely not be available for OSX. Right? 
> 

No.

Cheers,
Simon



More information about the R-devel mailing list