On 09/17/2013 01:39 AM, Ramon Diaz-Uriarte wrote:
Dear Martin,
I am not sure to fully understand (or understand the consequences)
mtmorgan at fhcrc.org writes:
Bioconductor developers:
The Bioc release schedule is available
http://bioconductor.org/developers/release-schedule/
A reminder that version 2.13 of Bioconductor will be released on October 15, and
will work with R-3.0.2. Package authors should be testing the Bioc 'devel'
version of their package with R-3.0.2 pre-release.
R-3.0.2 is scheduled for release on September 25. The lack of synchrony between
R and Bioc releases introduces a wrinkle. Users of R-3.0.2 (including the
pre-release, which became available earlier today) using install.packages()
(rather than the recommended biocLite()) will be directed toward Bioconductor
2.13 packages, so will be getting a preview of the Bioconductor release
scheduled for October 15.
So the recommended way (for now) to test our packages is to get any possible
BioC dependencies using install.packages (after setting the repository with,
say, biocinstallRepos() )? So the mechanism in
http://www.bioconductor.org/developers/useDevel/
followed by biocLite() is not recommended?
Unfortunately, the scheduling requirements of our own release cycle (including
building annotation packages and providing fair warning for new package
contributors) did not allow us to move the Bioconductor release forward. The
chosen solution seemed least disruptive of the available options.
Actually, R-3.0.1 _Patched_, from 2013-09-03 r63824.
The R used in bioc-devel tracks approximately the R which users will eventually
experience when bioc-devel becomes release 2.13. The R used for bioc-LATEST is
updated periodically and will be updated (to the pre-release version of R-3.0.2)
'soon' (perhaps this has already been done, and we'll see when the build reports
appear at around 10am today).
b) Suppose package A depends on package B. Package B (with version as available
in BioC devel, and with tests as shown in
http://www.bioconductor.org/checkResults/2.13/bioc-LATEST/) is working OK with
R-3.0.1. However, package B now breaks under R-3.0.2, or generates new warning
(which would mean it is not acceptable for BioC). What will the developer of A
get when she or he tries to install the dependencies of A? And what should the
developer of A do?
The packages should be passing build and check under the latest version of
R-3.0.2 available.
c) What will be the policy of BioC regarding the new notes that R-3.0.2 might
generate (and that have been the source of several threads in the R-devel list
in the last few weeks)? I've re-read
http://www.bioconductor.org/developers/package-guidelines/
and I can only find "Authors should also try to address all notes that arise
during build or check."
Almost always notes point to good programming practice. This applies to more
recent notes as much as it does to previously existing notes and warnings. The
long threads on R-devel have been interspersed with valuable insights into,
e.g., how to correctly place packages in to Imports:, how to use importFrom and
export, and when to use :: (and when not to use :::). It is almost always in the
developer's best interest to pay attention to the notes, warnings, and errors,
and Bioconductor expects our developers to act accordingly
Martin
R-3.0.2 allows for an environment variable to influence the version of
Bioconductor in use, from the R NEWS file:
? The default ?version? of Bioconductor for its packages has been
changed to the upcoming 2.13, but this can be set by the
environment variable R_BIOC_VERSION, e.g. in file Renviron.site.
Martin Morgan,
Bioconductor