Feature Freeze Exception Request for r-base 2.9.2

This is a feature freeze exception request to integrate support for the Revolution R packages into Ubuntu 9.10. We've been planning to support Revolution's R alongside Gnu R in 9.10 since UDS, however getting the packaging details just right for the two to coexist took longer than we thought to do right.

I believe the packages are now ready, however.

This also technically requires a slight upstream version increase (2.9.1 to 2.9.2). Few features are added:

CHANGES IN R VERSION 2.9.2

NEW FEATURES

    o install.packages(NULL) now lists packages only once even if they
        occur in more than one repository (as the latest compatible
        version of those available will always be downloaded).

    o approxfun() and approx() now accept a 'rule' of length two, for
        easy specification of different interpolation rules on left and
        right.

        They no longer segfault for invalid zero-length specification
        of 'yleft, 'yright', or 'f'.

    o seq_along(x) is now equivalent to seq_len(length(x)) even where
        length() has an S3/S4 method; previously it (intentionally)
        always used the default method for length().

    o PCRE has been updated to version 7.9 (for bug fixes).

    o agrep() uses 64-bit ints where available on 32-bit platforms
        and so may do a better job with complex matches.
        (E.g. PR#13789, which failed only on 32-bit systems.)

DEPRECATED & DEFUNCT

    o R CMD Rd2txt is deprecated, and will be removed in 2.10.0.
        (It is just a wrapper for R CMD Rdconv -t txt.)

    o tools::Rd_parse() is deprecated and will be removed in 2.10.0
        (which will use only Rd version 2).

BUG FIXES

    o parse_Rd() still did not handle source reference encodings
        properly.

    o The C utility function PrintValue no longer attempts to print
        attributes for CHARSXPs as those attributes are used
        internally for the CHARSXP cache. This fixes a segfault when
        calling it on a CHARSXP from C code.

    o PDF graphics output was producing two instances of anything
        drawn with the symbol font face. (Report from Baptiste Auguie.)

    o length(x) <- newval and grep() could cause memory corruption.
        (PR#13837)

    o If model.matrix() was given too large a model, it could crash
        R. (PR#13838, fix found by Olaf Mersmann.)

    o gzcon() (used by load()) would re-open an open connection,
        leaking a file descriptor each time. (PR#13841)

    o The checks for inconsistent inheritance reported by setClass()
        now detect inconsistent superclasses and give better warning
        messages.

    o print.anova() failed to recognize the column labelled
        P(>|Chi|) from a Poisson/binomial GLM anova as a p-value
        column in order to format it appropriately (and as a
        consequence it gave no significance stars).

    o A missing PROTECT caused rare segfaults during calls to
        load(). (PR#13880, fix found by Bill Dunlap.)

    o gsub() in a non-UTF-8 locale with a marked UTF-8 input
        could in rare circumstances overrun a buffer and so segfault.

    o R CMD Rdconv --version was not working correctly.

    o Missing PROTECTs in nlm() caused "random" errors. (PR#13381 by
        Adam D.I. Kramer, analysis and suggested fix by Bill Dunlap.)

    o Some extreme cases of pbeta(log.p = TRUE) are more accurate
        (finite values < -700 rather than -Inf). (PR#13786)

        pbeta() now reports on more cases where the asymptotic
        expansions lose accuracy (the underlying TOMS708 C code was
        ignoring some of these, including the PR#13786 example).

    o new.env(hash = TRUE, size = NA) now works the way it has been
        documented to for a long time.

    o tcltk::tk_choose.files(multi = TRUE) produces better-formatted
        output with filenames containing spaces. (PR#13875)

    o R CMD check --use-valgrind did not run valgrind on the package
tests.

    o The tclvalue() and the print() and as.xxx methods for class
        "tclObj" crashed R with an invalid object -- seen with an
        object saved from an earlier session.

    o R CMD BATCH garbled options -d <debugger> (useful for
        valgrind, although --debugger=valgrind always worked)

    o INSTALL with LazyData and Encoding declared in DESCRIPTION
        might have left options("encoding") set for the rest of the
        package installation.

The package has been tested and living in the RevoR ppa: https://edge.launchpad.net/~revor/+archive/ppa -- an install and build log is available there.