further on ESS/XEmacs for Windows

[Moved to R-devel, as this is going to get technical and is already a
minority interest.]

"john" == John Fox <jfox@mcmaster.ca> writes:

    john> (2) I spent a good part of the day today trying XEmacs and ESS on a
    john> variety of Windows systems -- with and without my configuration
    john> files. I noticed a number of intermittent problems with GUI-related
    john> commands, such as attempts to open dialog boxes causing the R process
    john> to freeze. I'll try to track down the source of these problems.

I'm not sure what the status of these is -- I do know that some people
I trust have wondered how the heck R even works under (X)Emacs on
Windows (issues arising with systems interfacing, blocking calls, and
similar "stuff").

That's right!  rterm used directly uses a separate thread for input to
avoid it blocking (not processing Windows events) whilst waiting for
input.  rterm used by ESS is effectively used in batch mode: it was never
the intention that the GUI interactions should work.

Guido and I once (in 2000) discussed a way out using a supervisor thread,
and Guido implemented it (although I never saw the code),  He said it
worked under NT but not Win98.  That's the real problem -- supporting
low-end versions of Windows which do not behave as documented, when the
developers (then) were only using NT/2000.  On Unix one uses select() and
since we have since made that work to support url/socket connections it
would be possible to try to use that.  However, it would be a lot of work
to provide something that the main developers would never use, and we
remain desperately short of contributions by users of R on Windows.

Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595