Hi, I'm developing an R/TclTk application which uses the BWidget and Tktable Tk extensions and I'm trying to make it easy to install. For now, I'm focusing on Windows users who start with nothing. I have built Tcl/Tk and Tktable for windows using Msys/MinGW. (BWidget contains only Tcl scripts so does not need building.) I have then copied Tktable and BWidget into the lib subdirectory of the Tcl directory in R source, created by unzipping R_Tcl.zip from Brian Ripley's webpage. (I should also try creating the entire R_Tcl.zip myself, rather than mixing and matching.) Finally I have followed the instructions in src/gnuwin32/INSTALL including "make distribution" which has built an R installation wizard including the Tk extensions. It seems to work fine and I don't think I've violated any license agreements - it's my understanding that Tktable and BWidget source (both on sourceforge) are under BSD, like the other Tcl already included with R. Of course some would say that you can build faster and smaller Tcl libraries using Visual C++. I didn't notice much difference with my application running Tktable. I did try buildling Tcl/Tk with the free version of VC++.NET SDK but nmake -f makefile.vc in Tcl_source/win/ complained that vcvars32.bat needed to be run. The latest version of the (free) VC++.NET SDK seems to only have sdkvars.bat, which does something similar to vcvars32.bat (i.e. set PATH, INCLUDE, LIB and NetSamplePath), but running it didn't seem to help so I turned to Msys/MinGW. Maybe I should buy the full Visual C++/Visual Studio? I'm wondering whether anyone would like to point out any potential problems (license or otherwise) with distributing a customized R installation wizard. Also, I would be happy to become involved in the maintenance/testing of R_Tcl.zip for windows if I'm qualified, but I suppose this would require buying the full Visual C++/Visual Studio, right? If there is sufficient building/testing from volunteers, are the R core developers considering bundling more Tcl/Tk packages with R, or does the problem of HOW MANY packages/megabytes to add mean that there would always have to be a separate "batteries included" installation wizard? Regards, James
Have you tried to make R under MSYS/MinGW, there would seem to be some obvious benefits to this over the current collection of Cygwin tools? In the end I don't think a separate windows install of R should be needed... in much the same way as the separate apple mac install has been discontinued in 1.8.0 now they have the osX/fink setup. I've had a play at configuring and making R with MSYS and I seem to be able to get quite a long way down the process, but I'd be interested in your views since you obviously have much more experience in this area. Regards, John Marsland> -----Original Message----- > From: wettenhall@wehi.EDU.AU [mailto:wettenhall@wehi.EDU.AU] > Sent: 30 September 2003 08:42 > To: r-devel@stat.math.ethz.ch > Subject: [Rd] Adding Tk extensions to R for windows > > > Hi, > > I'm developing an R/TclTk application which uses the BWidget and > Tktable Tk extensions and I'm trying to make it easy to install. > For now, I'm focusing on Windows users who start with nothing. > > I have built Tcl/Tk and Tktable for windows using Msys/MinGW. > (BWidget contains only Tcl scripts so does not need building.) > I have then copied Tktable and BWidget into the lib subdirectory > of the Tcl directory in R source, created by unzipping R_Tcl.zip > from Brian Ripley's webpage. (I should also try creating the > entire R_Tcl.zip myself, rather than mixing and matching.) > Finally I have followed the instructions in src/gnuwin32/INSTALL > including "make distribution" which has built > an R installation wizard including the Tk extensions. > > It seems to work fine and I don't think I've violated any > license agreements - it's my understanding that Tktable and > BWidget source (both on sourceforge) are under BSD, like the other > Tcl already included with R. > > Of course some would say that you can build faster and smaller Tcl > libraries using Visual C++. I didn't notice much difference with my > application running Tktable. I did try buildling Tcl/Tk with the > free version of VC++.NET SDK but nmake -f makefile.vc in > Tcl_source/win/ complained that vcvars32.bat needed to be run. The > latest version of the (free) VC++.NET SDK seems to only have > sdkvars.bat, > which does something similar to vcvars32.bat (i.e. set PATH, INCLUDE, > LIB and NetSamplePath), but running it didn't seem to help so I turned > to Msys/MinGW. Maybe I should buy the full Visual C++/Visual Studio? > > I'm wondering whether anyone would like to point out any potential > problems (license or otherwise) with distributing a customized R > installation wizard. > > Also, I would be happy to become involved in the maintenance/testing > of R_Tcl.zip for windows if I'm qualified, but I suppose this would > require buying the full Visual C++/Visual Studio, right? > > If there is sufficient building/testing from volunteers, are > the R core > developers considering bundling more Tcl/Tk packages with R, > or does the > problem of HOW MANY packages/megabytes to add mean that there > would always > have to be a separate "batteries included" installation wizard? > > Regards, > James > > ______________________________________________ > R-devel@stat.math.ethz.ch mailing list > https://www.stat.math.ethz.ch/mailman/listinfo/r-devel >********************************************************************** This is a commercial communication from Commerzbank AG.\ \ T...{{dropped}}
My understanding is that MSYS was designed to provide a Minimal SYStem to complement the Minimal GNU tools. So you can run the ./configure on windows just the same as any unix variant and make as normal. I understand it uses native libraries rather then the cygwin.dll .. so should be faster??? This would completely remove the need for separate windows project (except for any platform specific devices eg graphics, GUIs etc ... MSYS provides extensions for the win32 API also) - in exactly the same way and the new Mac osX/fink build works. One of my main reasons for interest is that I was trying to use the GTK/Glade libraries and it seemed as if quite a lot of modification was necessary for win32 systems. I got more interested when I started thinking about Rpvm and all the spare win32 boxes sitting around - curently this package only works for unix/linux systems. My guess is that an awful lot more of the R project could be made non-platform-specific if this route is followed. Or in their own words .... [http://www.mingw.org/msys.shtml] "This package provides a minimal POSIX system to allow a typical configure script to execute so that you can execute make. It provides a native mingw32 host/build/target system in order to use the MinGW port of GCC and binutils to port other packages to the Windows environment. It eliminates the complexities of the semi-cross build system provided by -mno-cygwin. It removes the need to use programs such as cygpath by auto translating the path strings appropriately. With version 1.0.8 both paths in arguments and paths in the environment are converted." The MSYS developer took kit provides the following packages: autoconf automake libtool autogen openssl openssh cvs guile inetutils Regards, John Marsland> -----Original Message----- > From: Duncan Murdoch [mailto:dmurdoch@pair.com] > Sent: 30 September 2003 12:56 > To: Marsland, John > Cc: r-devel@stat.math.ethz.ch > Subject: Re: [Rd] Adding Tk extensions to R for windows > > > On Tue, 30 Sep 2003 11:34:27 +0100, you wrote: > > >Have you tried to make R under MSYS/MinGW, there would seem > to be some > >obvious benefits to this over the current collection of > Cygwin tools? > > I haven't tried that; what benefits do you see? > > Duncan Murdoch >********************************************************************** This is a commercial communication from Commerzbank AG.\ \ T...{{dropped}}
I'd be interested in the detail of what you linked to to get Rpvm working on Windows. I've had a go at this using both the source and the binaries on windows. The problem I had was I don't have rsh only Openssh which I managed to get working on linux (client & server) in minutes with pvm. The attraction of MSYS was partly that it includes ssh in the DTK, since I need a wholey open-source solution.> -----Original Message----- > From: Simon Urbanek [mailto:simon.urbanek@math.uni-augsburg.de] > Sent: 30 September 2003 18:43 > To: Marsland, John > Cc: 'Duncan Murdoch'; r-devel@stat.math.ethz.ch > Subject: Re: [Rd] Adding Tk extensions to R for windows > > > On Tuesday, September 30, 2003, at 03:42 PM, Marsland, John wrote: > > > necessary for win32 systems. I got more interested when I started > > thinking > > about Rpvm and all the spare win32 boxes sitting around - > curently this > > package only works for unix/linux systems. My guess is that > an awful > > lot > > more of the R project could be made non-platform-specific if this > > route is > > followed. > > FYI: Rpvm itself is fairly platform-independent. I had a quick glance > at rpvm on Win32 quite recently. It compiled fine, the only serious > problem I had was at link-time due to the fact that PVM itself has no > shared library that could be easily used with MinGW (talking > about the > official binary version). Rpvm needed hardly any touchups. Presumably > compiling PVM with MinGW should solve the issue and rpvm should work > almost out of the box. Actually this shows that R makes it > possible for > developers to 'accidentally' produce packages that work on > Windows even > though they were developed for unix machines ;). > > Cheers, > Simon >********************************************************************** This is a commercial communication from Commerzbank AG.\ \ T...{{dropped}}