R devel - Feb 2008 - Mac OS X 2.5.1 / reg-plot.R / tr (PR#10781)

In building R 2.6.2 from source for Mac OS X 2.5.1 (with
i686-apple-darwin9-gcc-4.0.1), 'make check' fails.  The failure is in
reg-plot.R, and occurs because Mac OS X's 'tr' command (invoked by
Rdiff
to strip carriage returns) regards the dagger sign in reg-plot.ps as an
illegal byte sequence.

I'm surprised that this doesn't seem to have been reported before, but
it was already present at least in R 2.6.1

Incidentally, couldn't the handling of line endings now be left to
svn:eol-style?

Andrew Runnalls

-- 
Dr Andrew Runnalls,
Computing Laboratory,
University of Kent,
CANTERBURY CT2 7NF, UK

Tel: (0)1227 823821

On 17/02/2008 8:35 AM, A.R.Runnalls at kent.ac.uk wrote:
> In building R 2.6.2 from source for Mac OS X 2.5.1 (with
> i686-apple-darwin9-gcc-4.0.1), 'make check' fails.  The failure is
in
> reg-plot.R, and occurs because Mac OS X's 'tr' command (invoked
by Rdiff
> to strip carriage returns) regards the dagger sign in reg-plot.ps as an
> illegal byte sequence.
> 
> I'm surprised that this doesn't seem to have been reported before,
but
> it was already present at least in R 2.6.1
> 
> Incidentally, couldn't the handling of line endings now be left to
> svn:eol-style?
svn:eol-style assumes that the machine which checks out the working copy 
is the same as the machine doing the editing.  You get really horrible 
messes when you do something like mounting a Unix drive on a PC, or 
emailing a file from one system to another, etc.  I've worked with it 
and without it, and I much prefer working without it.

Duncan Murdoch

Please do as we ask in both the FAQ and posting guide and check the 
current working copy of R, which does this differently.

On Sun, 17 Feb 2008, A.R.Runnalls at kent.ac.uk wrote:
> In building R 2.6.2 from source for Mac OS X 2.5.1 (with
> i686-apple-darwin9-gcc-4.0.1), 'make check' fails.  The failure is
in
> reg-plot.R, and occurs because Mac OS X's 'tr' command (invoked
by Rdiff
> to strip carriage returns) regards the dagger sign in reg-plot.ps as an
> illegal byte sequence.
Known for certain locales: we do ask you to report yours (part of the 'at 
a minimum' information requested in the posting guide).
> I'm surprised that this doesn't seem to have been reported before,
but
> it was already present at least in R 2.6.1
It has been, and fixed before!  From the svn logs

r44231 | ripley | 2008-01-29 12:45:00 +0000 (Tue, 29 Jan 2008) | 1 line

MacOS tr does not cope with Latin-1 chars in a UTF-8 locale
> Incidentally, couldn't the handling of line endings now be left to
> svn:eol-style?
Have you read the code concerned?  It says

## some packages ship .Rout.save with CRLF endings

and we have no control over what line-endings third-party packages ship 
with.

-- 
Brian D. Ripley,                  ripley at 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 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595