[Rd] Suspect error in windows version of 3.60

Martin Maechler m@ech|er @end|ng |rom @t@t@m@th@ethz@ch
Mon Apr 29 10:19:53 CEST 2019


>>>>> Duncan Murdoch 
>>>>>     on Sat, 27 Apr 2019 15:00:20 -0400 writes:

    > On 27/04/2019 2:46 p.m., Duncan Murdoch wrote:
    >> On 27/04/2019 10:30 a.m., Len Weil wrote:
    >>> Gentlemen,
    >>> 
    >>> I've observed a suspected bug in R-3.60 for Windows 10.
    >>> 
    >>> I am using the latest R version, but I have observed the same error in all
    >>> releases: alpha, beta RC and released.
    >>> 
    >>> Also the same issue occurs in the 32 and 64 bit GUI versions.  The program
    >>> header I invoke is:
    >>> 
    >>> 
    >>> R version 3.6.0 Patched (2019-04-26 r76431) -- "Planting of a Tree"
    >>> Copyright (C) 2019 The R Foundation for Statistical Computing
    >>> Platform: x86_64-w64-mingw32/x64 (64-bit)
    >>> 
    >>> Specifically, the menu selection Packages -> install packages fails.
    >>> 
    >>> When invoked, it simply echoes: utils:::menuInstallPkgs() and returns the
    >>> prompt.
    >>> 
    >>> 
    >>> I found the underlying code with the getAnywhere function.
    >>> 
    >>> It displays:
    >>> 
    >>> getAnywhere(menuInstallPkgs)
    >>> 
    >>> A single object matching 'menuInstallPkgs' was found
    >>> It was found in the following places
    >>> namespace:utils
    >>> with value
    >>> 
    >>> function (type = getOption("pkgType"))
    >>> {
    >>>   install.packages(NULL, .libPaths()[1L], dependencies = NA,
    >>>                    type = type)
    >>> }
    >>> 
    >>> <bytecode: 0x0000000016c6a730>
    >>> <environment: namespace:utils>
    >>> 
    >>> 
    >>> The first argument, NULL, prevents the underlying function install.packages
    >>> from working.
    >> 
    >> No, it is documented to trigger presentation of a list of packages to
    >> install.  That list takes time to download from CRAN, so perhaps you are
    >> just too impatient to wait for it, or perhaps it really is broken:  but
    >> the NULL is not the issue.
    >> 

    > Sorry, you were right.  The docs in 3.6.0 have changed:  now the pkgs 
    > argument must be missing to trigger the menu.  The NULL is not missing, 
    > it is interpreted as specifying no packages.

    > Duncan Murdoch

Thank you, Len and Duncan;
Duncan additionally for the formal R bug report with patch PR#17556 :
https://bugs.r-project.org/bugzilla/show_bug.cgi?id=17556

I cannot easily build R for windows from the sources either, but
committed Duncan's fix to the R-devel sources for now,
in svn rev 76434,  so we can all install a binary version of
R-devel for Windows (>= rev 76434) in a day or two
from CRAN https://cran.r-project.org/bin/windows/base/rdevel.html
thanks to Jeroen Ooms'  autobuilder.

If that confirms the problem fixed, we will of course port it to
R 3.6.0 patched, so it will be in R 3.6.1 (which is *not* scheduled yet).

Martin Maechler
ETH Zurich and R Core Team



More information about the R-devel mailing list