[R-sig-Debian] Fwd: RKWard and R 4.0.0 - Important bug and workaround

Johannes Ranke joh@nne@@r@nke @end|ng |rom jrwb@de
Sun May 24 14:49:14 CEST 2020


Dear all,

as I am providing rkward backports for Debian testing and stable on CRAN, I am 
forwarding this notice by the main developer.

Meanwhile, it has been communicated that R 4.0.1 will be released on June 6.

Cheers,

Johannes


----------  Weitergeleitete Nachricht  ----------

Betreff: RKWard and R 4.0.0 - Important bug and workaround
Datum: Sonntag, 24. Mai 2020, 10:14:17 CEST
Von: Thomas Friedrichsmeier <thomas.friedrichsmeier using ruhr-uni-bochum.de>
An: rkward-devel using kde.org <rkward-devel using kde.org>, rkward-users using kde.org

Hi!

This is a note of caution about using RKWard with R 4.0.0:

-- Problem --

There is a bug in R 4.0.0 with respect to active bindings, which RKWard
is relatively likely to trigger. Running the following code *twice*
will render your RKWard session unusable, and **could result in
loss of data**:

  for (i in 1:2) { print (i) }

Note that this affects top-level statements, only, i.e. in general code
that you enter in the console of run from a script, directly. Not code
in packages or RKWard plugins.

-- Workaround --

If you have the choice, please delay upgrading R until R 4.0.1 is
released.

If you have already installed R 4.0.0, you should set the option

  compiler::enableJIT(2)  # or a lower value

in each session. This can also be made permanent by copying it into
the field "Futher option commands to run in each session" in
Settings->Configure RKWard->R-Backend.

-- Outlook --

The bug has been fixed in R devel, but no release date has been set to
R 4.0.1, yet. I will probably create a "bugfix"-release that enables
the above workaround by default in the coming day. Meanwhile I have
started working on a new approach to object modification detection that
does not rely on active bindings.

Regards
Thomas



More information about the R-SIG-Debian mailing list