[R-pkg-devel] R CMD build with configure-args

[Duncan Murdoch]

On 30/07/2015 4:52 PM, Russell Almond wrote:
> The behavior changed somewhere along the line. 

If you don't want to spend the time to answer a simple question like
"when did you see that?", why do you expect others to spend time solving
your problem?

Duncan Murdoch

 I think it is related to
> the version bump from 3.1 to 3.2, but it could have been earlier.
> 
> It used to be when I wanted to release a version of my package I would run:
> 
> INSTALL
> check
> build
> INSTALL --build (on each of a Windows and Mac box).
> 
> to create source and binary tarballs.
> 
> I was never quite sure if I needed the check syntax or the INSTALL 
> syntax for build.  Somewhere along the line it stopped working.  I'm not 
> sure exactly when, because I was working around by creating the source 
> tarball on the Windows box where I was using an environmental variable 
> instead of the command line arg.
> 
> I suspect that what must have happened is that previously build did not 
> try to recompile the package (or used my source directory where the 
> binaries where already in place).  Now it does, so it would probably be 
> useful to add the --install-args switch.
> 
> 	--Russell
> 
> 
> On 07/30/2015 04:08 PM, Duncan Murdoch wrote:
>> On 30/07/2015 2:53 PM, Russell Almond wrote:
>>>
>>> I've got a package that uses C code to link R to a third-party library.
>>>
>>> I've set up my configure script to take an argument --with-netica
>>> which passes the location of the library, header files, &c.
>>>
>>> So I can install the package using:
>>>
>>>     $ R CMD INSTALL RNetica
>>> --configure-args='--with-netica=/home/ralmond/software/Netica_API_504/'
>>>
>>> and check the packge (build the manual) using:
>>>
>>>     $ R CMD check RNetica
>>> --install-args="--configure-args='--with-netica=/home/ralmond/software/Netica_API_504/'"
>>>
>>> My problem comes when I want to build a source tarball for
>>> distribution.  I used to use:
>>>
>>>     $ R CMD build RNetica
>>> --install-args="--configure-args='--with-netica=/home/ralmond/software/Netica_API_504/'"
>>
>> When you say "used to use", when was that?  There was no --install-args
>> option in R 3.0.  I think there should be, and I can add it, but I'd
>> like to know if it was removed for a reason.
>>
>> Duncan Murdoch
>>
>>>
>>> but that generates an error:
>>>
>>>     gcc -std=gnu99 -I/usr/share/R/include -DNDEBUG     -I. -I/src -fpic
>>> -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat
>>> -Werror=format-security -D_FORTIFY_SOURCE=2 -g  -c Cases.c -o Cases.o
>>>     In file included from Cases.c:18:0:
>>>     ./RNetica.h:4:20: fatal error: Netica.h: No such file or directory
>>>      #include <Netica.h>
>>>                         ^
>>>
>>> basically, it the Includes directory is not getting set correctly.  If
>>> I use the same syntax as the INSTALL (i.e., just configure-args and
>>> not the install-args wrapper) I get the same error message.  R CMD
>>> INSTALL --build creates a binary tarball and not a source.
>>>
>>> I think used to work correctly under R 3.1.  Did I miss something in
>>> the release notes?
>>>
>>> I'm having this problem under Ubuntu 14.4 (I think I'm also having the
>>> problem on Mac OS X; I'm not having the problem under Windows, but
>>> mostly because I'm using an environmental variable instead of a
>>> command line argument to pass the location of the 3rd party library.)
>>>
>>> Thanks in advance for any help.
>>>
>>
>