[Rd] Bug in versioned install (was: (fwd) R-1.7.0 : Problem
withDownloading "dse") (PR#2827)
murdoch at stats.uwo.ca
Fri Apr 25 14:37:31 MEST 2003
On Fri, 25 Apr 2003 18:21:17 +0200, you wrote in message
<001701c30b46$b310e130$e502eb82 at alpha.wehi.edu.au>:
>A patch for install.packages() can be loaded as
>All fixes are done to the local function unpackPkg() of
>install.packages(). The fix is based on the assumption that a bundle (in
>zip format) contains directories for packages , which in turn contains
>the DESCRIPTION files needed. Is this assumption correct?! I also made
>the temporary directory to be deleted using on.exit() to make the error
>recovering a little bit nicer.
That won't quite work. There are two problems.
- With namespaces any references from within base functions to
install.packages will see the old function. There are a few other
functions that call install.packages, so they would need to be fixed,
and the functions that call them, etc... It's best to fix the
original definition of install.packages by fixing the base package.
- A change was needed to the code that builds packages as well as to
install.packages. Binary bundles built with the new code will include
a DESCRIPTION file for the bundle (as well as separate files for each
package within it). Without that file, the new install.packages won't
As mentioned the other day, I've put a build of the current patched
version of R on my web site (at
http://www.stats.uwo.ca/faculty/murdoch/software/r-devel). I expect
there will be a 1.7.1 release, but a release date for it has not been
set. You can use the one from my web site in the meantime if you need
any of the recent bug fixes.
More information about the R-devel