Hello Maintainers, I just wanted to mention, that the new GUI (ref 3141), which was recommended as an update when updating to R 2.3.0 still exhibits some undesired behaviour. So i can't save images from the quartz device as pdf anymore. The programme wants to save as Quartz image. So i can do this and then later open the file in preview and saving it as a pdf. but before it was much more convenient. Maybe it is a general problem with the Quartz devices: Often i got two of them automatically and one is always empty and it's not possible to get rid of it. This doesn't cause any further problem it is just a bit nagging. However, i greatly appreciate your effort of bringing R to Aqua, and i am happy that i finally found R and learned (beginner stage) how to use it. One wish: The status region at the bottom of the console window could be a bit bigger (or growing, to always display the whole function usage) and the font used there could be a bit bigger (or adjustable). Thank you Gerald Jurasinski P.S. Does anybody know a programme or package capable of producing two-dimensional correlograms (Oden & Sokal 1986)? ?????????????????? gerald jurasinski biogeography university of bayreuth 95440 bayreuth gerald.jurasinski at uni-bayreuth.de
new GUI
7 messages · Gerald Jurasinski, Kieran Healy, mwtoews at sfu.ca +2 more
Hello, Sorry to bother you, but I have a problem installing R-2.3.0 from source. 'configure' and 'make' produce no error, and the R executable seems to run fine (particularly, R --version does work), but 'make install' fails with the following error message : % sudo make install Password: make[1]: Nothing to be done for `install'. make[1]: Nothing to be done for `install'. installing doc ... help2man: can't get `--version' info from ../bin/R make[1]: *** [R.1] Error 22 make: *** [install] Error 1 R-2.2.1 compiles and installs without problem. Using MacOS 10.4.6 on a powerMac G5 with XCode 2.2.1. Thank you, Jean
Jean Thioulouse - Labo de Biometrie et Biologie Evolutive, UMR CNRS 5558
Universite Lyon 1, Batiment G. Mendel, 43 Boulevard du 11 Novembre 1918
69622 Villeurbanne Cedex, France. Tel/Fax : (33) 4 72 43 27 56
http://pbil.univ-lyon1.fr/JTHome.html
Hi - I posted a little while ago about the same problem, as did Michael Toews. I've tried to follow this up. The short answer is that R 2.3.0's version string doesn't conform to what the help2man.pl utility expects. The help2man.pl script gets version information from R --version and makes a man page from it. Here is the relevant line of the Makefile inside the doc/ directory of the R-2.3.0 source code bundle:
@$(HELP2MAN) --include=$(srcdir)/R.aux --no-info --output=R.1 \
--name="a language for data analysis and graphics" \
$(top_builddir)/bin/R
The help2man script wants the version information in a specific form Here is 2.3.0's version info (without the copyright message, which doesn't matter for the error):
[Deiseal] ~/downloads/R-2.3.0/doc % ../bin/R --version R version 2.3.0 (2006-04-24)
And here is 2.2.1:
[Deiseal] ~ % R --version R 2.2.1 (2005-12-20).
The extra word 'version' is what does it. The roots of this issue are in src/main/version.c, which contains the code that produces the text of the version message. It has changed since 2.2.1. The changed code includes a function whose comments say, more or less, "this is a workaround designed to make GNU tools like help2man happy," together with information about how the version message will change in the next version of R. So I guess the authors of the code are aware of this issue, but for whatever reason it still arises on some macs. I'm afraid I don't' know C, so I can't diagnose any further. As a short-term, completely ad hoc workaround, I ran the help2man command in the Makefile directly, but pointed it at my already- installed version of R (2.2.1) instead. I.e., from inside the doc/ directory in the R 2.3.0 sources, and assuming you have R 2.2.1 already installed, do
perl ../tools/help2man.pl --include=./R.aux --no-info --output=R.1 --name="a language for data analysis and graphics" R
(This should all be one line.) This produces a file called R.1 containing version information, which is the desired goal, but with version info from R.2.2.1 instead of 2.3.0. You can edit the first line of the file to make the version number correct if you like. Then you can comment out the equivalent section in doc/Makefile, viz,
#@$(HELP2MAN) --include=$(srcdir)/R.aux --no-info --output=R.1 \ # --name="a language for data analysis and graphics" \ # $(top_builddir)/bin/R
Doing 'sudo make install' as usual (from the top-level source directory) will then work, as the R.1 file is now present in the doc directory. Of course this doesn't change the usual output of R -- version (you'd have to edit version.c to do that), just the information in the man page. Obviously is not any kind of proper solution, but I'm afraid I don't know C, or the ins-and-outs of R's source code in great detail. I'm not sure why this issue only appears on some macs (e.g., my iBook) and not others (e.g., by G5 tower). Best, Kieran
On May 16, 2006, at 1:32 AM, Jean Thioulouse wrote:
Hello,
Sorry to bother you, but I have a problem installing R-2.3.0 from
source.
'configure' and 'make' produce no error, and the R executable seems
to run
fine (particularly, R --version does work), but 'make install'
fails with
the following error message :
% sudo make install
Password:
make[1]: Nothing to be done for `install'.
make[1]: Nothing to be done for `install'.
installing doc ...
help2man: can't get `--version' info from ../bin/R
make[1]: *** [R.1] Error 22
make: *** [install] Error 1
R-2.2.1 compiles and installs without problem. Using MacOS 10.4.6 on a
powerMac G5 with XCode 2.2.1.
Thank you,
Jean
--
Jean Thioulouse - Labo de Biometrie et Biologie Evolutive, UMR CNRS
5558
Universite Lyon 1, Batiment G. Mendel, 43 Boulevard du 11 Novembre
1918
69622 Villeurbanne Cedex, France. Tel/Fax : (33) 4 72 43
27 56
http://pbil.univ-lyon1.fr/JTHome.html
_______________________________________________ R-SIG-Mac mailing list R-SIG-Mac at stat.math.ethz.ch https://stat.ethz.ch/mailman/listinfo/r-sig-mac
Hi,
The problem is actually R's inability to print the version number
normally .. for instance typing: './bin/R --version' produces the
proper output, but './bin/R --version | cat' does not produce any
output ('help2man' reads an empty strings, and throws an error).
The only solution for now is to download a pre-compiled version from
CRAN, since it works perfectly (including 'R --version | cat').
However, I have tracked down the vulnerability to:
sys-std.c: In function 'pushReadline':
sys-std.c:497: warning: passing argument 2 of
'rl_callback_handler_install' from incompatible pointer type
sys-std.c: In function 'popReadline':
sys-std.c:513: warning: passing argument 2 of
'rl_callback_handler_install' from incompatible pointer type
The source file ./src/unix/sys-std.c refers to Apple's ReadLine on OS
X 10.4-pre, and how it needs to be forced to flush the stdout to make
it visible (this is exactly the problem described above from
help2man). Simon fixed this almost a year ago (svn r33778), but it
appears this problem has re-surfaced for some proportions of Macs.
I can provide a complete log from './configure', 'make' and 'sudo
make install' to anyone interested in seeing.
+mt
On 2006-May-16, at 01:32 , Jean Thioulouse wrote:
Hello,
Sorry to bother you, but I have a problem installing R-2.3.0 from
source.
'configure' and 'make' produce no error, and the R executable seems
to run
fine (particularly, R --version does work), but 'make install'
fails with
the following error message :
% sudo make install
Password:
make[1]: Nothing to be done for `install'.
make[1]: Nothing to be done for `install'.
installing doc ...
help2man: can't get `--version' info from ../bin/R
make[1]: *** [R.1] Error 22
make: *** [install] Error 1
R-2.2.1 compiles and installs without problem. Using MacOS 10.4.6 on a
powerMac G5 with XCode 2.2.1.
Thank you,
Jean
--
Jean Thioulouse - Labo de Biometrie et Biologie Evolutive, UMR CNRS
5558
Universite Lyon 1, Batiment G. Mendel, 43 Boulevard du 11 Novembre
1918
69622 Villeurbanne Cedex, France. Tel/Fax : (33) 4 72 43
27 56
http://pbil.univ-lyon1.fr/JTHome.html
_______________________________________________ R-SIG-Mac mailing list R-SIG-Mac at stat.math.ethz.ch https://stat.ethz.ch/mailman/listinfo/r-sig-mac
Michael, thanks for tracking this down, excellent work! The reason for CRAN R binary working is simply that it uses GNU readline instead of Apple's libedit (which is disguised as readline btw). A simple work-around is therefore to install GNU readline (see the R for Mac OS X FAQ for details, don't forget -DNEED_EXTERN_PC). I think I may be able to reproduce this on my laptop, I'll try that tomorrow and see if I can fix it. Cheers, Simon
On May 16, 2006, at 12:27 PM, Michael Toews wrote:
Hi,
The problem is actually R's inability to print the version number
normally .. for instance typing: './bin/R --version' produces the
proper output, but './bin/R --version | cat' does not produce any
output ('help2man' reads an empty strings, and throws an error).
The only solution for now is to download a pre-compiled version from
CRAN, since it works perfectly (including 'R --version | cat').
However, I have tracked down the vulnerability to:
sys-std.c: In function 'pushReadline':
sys-std.c:497: warning: passing argument 2 of
'rl_callback_handler_install' from incompatible pointer type
sys-std.c: In function 'popReadline':
sys-std.c:513: warning: passing argument 2 of
'rl_callback_handler_install' from incompatible pointer type
[...]
Simon, Thanks for the message. I installed GNU readline 5.1, but i still get the help2man error. Jean
At 22:12 -0400 16/05/06, Simon Urbanek wrote:
Michael, thanks for tracking this down, excellent work! The reason for CRAN R binary working is simply that it uses GNU readline instead of Apple's libedit (which is disguised as readline btw). A simple work-around is therefore to install GNU readline (see the R for Mac OS X FAQ for details, don't forget -DNEED_EXTERN_PC). I think I may be able to reproduce this on my laptop, I'll try that tomorrow and see if I can fix it. Cheers, Simon
Jean Thioulouse - Labo de Biometrie et Biologie Evolutive, UMR CNRS 5558
Universite Lyon 1, Batiment G. Mendel, 43 Boulevard du 11 Novembre 1918
69622 Villeurbanne Cedex, France. Tel/Fax : (33) 4 72 43 27 56
http://pbil.univ-lyon1.fr/JTHome.html
Hi Kieran, This works for me, thanks. It will allow me to wait until Simon can fix the problem (hopefully). Jean
At 6:45 -0700 16/05/06, Kieran Healy wrote:
As a short-term, completely ad hoc workaround, I ran the help2man command in the Makefile directly, but pointed it at my already- installed version of R (2.2.1) instead. I.e., from inside the doc/ directory in the R 2.3.0 sources, and assuming you have R 2.2.1 already installed, do
perl ../tools/help2man.pl --include=./R.aux --no-info --output=R.1 --name="a language for data analysis and graphics" R
(This should all be one line.) This produces a file called R.1 containing version information, which is the desired goal, but with version info from R.2.2.1 instead of 2.3.0. You can edit the first line of the file to make the version number correct if you like. Then you can comment out the equivalent section in doc/Makefile, viz,
#@$(HELP2MAN) --include=$(srcdir)/R.aux --no-info --output=R.1 \ # --name="a language for data analysis and graphics" \ # $(top_builddir)/bin/R
Doing 'sudo make install' as usual (from the top-level source directory) will then work, as the R.1 file is now present in the doc directory. Of course this doesn't change the usual output of R -- version (you'd have to edit version.c to do that), just the information in the man page.
Jean Thioulouse - Labo de Biometrie et Biologie Evolutive, UMR CNRS 5558
Universite Lyon 1, Batiment G. Mendel, 43 Boulevard du 11 Novembre 1918
69622 Villeurbanne Cedex, France. Tel/Fax : (33) 4 72 43 27 56
http://pbil.univ-lyon1.fr/JTHome.html