[ESS-bugs] ESS 13.09-1 with old R versions fails pretty badly

[Vitalie Spinu]

 >>> Martin Maechler on Mon, 28 Oct 2013 13:30:59 +0100 wrote:

 > As an R core developer,
 > I have many old versions of R installed, that I occasionally use
 > for testing, comparison etc.

 > I just found that this is quite a problem with the new ESSR (or
 > other?) behavior of ESS. Of course ESSR is nowadays installed for me
 > (in a private library), but the installation from R 3.0.2 cannot be
 > used in R 2.1.1 (e.g.) :

 >> require("ESSR")
 > Loading required package: ESSR
 > Error in eval(expr, envir, enclos) : couldn't find function ".getNamespace"
 > In addition: Warning message:
 > package 'ESSR' was built under R version 3.0.2 
 > Error: unable to load R code in package 'ESSR'
 > [1] FALSE
 >> 

As a side note. Most other packages installed under 3.0.2 will probably
not work in 2.1.1.

 > For that reason, ESS trys to call install.packages(...)
 > but fails with that as well ...

 > One reason that you now name the package tarball  ESSR.tar.gz
 > instead of  ESSR_<ver>.tar.gz

Indeed ... it is not wise to keep all ESS_ver.tar.gz in svn source.

 > The latter is required for old(er) versions of R.

 > But even then, problems don't end.
 > The R code can't even be *parse*d because it uses '0L' which did
 > not exist in R 2.1.1 (e.g.).... actually I want to use even
 > older versions of R.

 > Of course, we can argue that "nobody uses such old versions of R",
 > but as you see it's not quite true, and the problem really that
 > ``almost nothing works'' now when ESSR is not available, not even
 > simple function name TAB completion, or even '?' fails because
 > .ess_help is undefined.


Versions older than 2.8 will probably not work correctly because of the
lack of completion or name-space support. Users (developers) who
occasionally want to load such an old R will have to do it in terminal
or emacs shell or whatever.

 > I think we need a workaround for old R versions such that a bit more
 > things work than now...

Sorry, but I have no time, nor desire, to work on that for the same
reason I don't work on emacs 22 compatibility. If anybody is willing to
contribute that feature, no problem on my side. But remember that it
might get quite stuffy.

We probably should add a note to the next release that R older than 2.8
is not supported.

   Vitalie