Arjen, Arjen Markus wrote on 2013-06-26: ---------------- > Note, when I wrote "minimal distribution" I was thinking more along > the lines of minimzing the amount of work to build the distribution.
Understood, but these two are not completely unrelated concepts, or at least they minimise different burdens on the user. I think it would be straightforward to create a minimal version of plplot that could be easily built by anyone with a gcc compiler. It would drive limited devices, e.g., ps/psc, svg, wingcc (windows) and xwin (*nix). It would have limited bindings (c, c++, fortran). And it would build from a single flat src directory and a single relatively short Makefile (probably different for each target platform to cope with pathnames and such) with a few editable lines at the top - or as we do for QSAS an editable shell script that sets up environment variables and calls make. This minimises the work in getting something up and running. You have a better feel than I do concerning the plplot user community to gauge what percentage of your users might have been happy to start this way. Deconstructing the full cmake plplot distribution by then, say, adding separate bundles to add other bindings or devices may or may not make sense depending on the user community (i.e. whether one or two of these wo uld reach a large fraction of the community or enable the plplot community to grow faster). This is a tradeoff between user effort and development effort, of course. Cmake, configure, scons and such are designed to keep a single development tree to ease maintenance, avoid duplication, and automagically solve platform/operating system quirks. In my experience, when they work, they're great; when they don't, many end-users will give up. Binary distributions are great for many end-users because they run out of the box (assuming they run - otherwise again end-users will give up). My impression is that the most difficult binaries are in the *nix world. It seems easier to create windows and mac binaries that will run without being too fussy about the version of everything else that is installed, but I have less experience in compiling against distributed 3rd party binary libraries which is what any plplot user would need to do. I wonder if a fairly drastic re-write of the windows (and other?) build/installation instructions for plplot would have been sufficient to help the original poster - Clive Page - make progress. In my experience, those instructions need to include step-by-step guidance about installing any other packages (cmake, mingw, msys, qt, ...) as well as how to build plplot. It should also give guidance about essential (e.g., gcc, say) vs optional (qt, wxWidgets, ...) requirements. Getting to the point where the plplot equivalent of "Hello world" compiles and runs provides an enormous encouragement to the new user. You may recall some time ago I suggested an example zero which would take the user through the steps to compile and run their own first plplot program, as opposed to the multitude of (very fine and comprehensive) plplot examples that get built by including special plcdemos.h headers and via the cmake build of all of plplot. There is such an example (x00) but not, as far as I can see, instructions about how a user can copy that to their own directory and compile it. Best wishes Steve -------------------------------------------------------------------- Professor Steven J Schwartz Phone: +44 (0)207 594 7660 Head, Space & Atmospheric Physics Fax: +44 (0)207 594 7772 The Blackett Laboratory Email: s.schwa...@imperial.ac.uk Imperial College London Office: Huxley 6M67A London SW7 2AZ, UK Web: www.sp.ph.ic.ac.uk/~sjs -------------------------------------------------------------------- ------------------------------------------------------------------------------ This SF.net email is sponsored by Windows: Build for Windows Store. http://p.sf.net/sfu/windows-dev2dev _______________________________________________ Plplot-devel mailing list Plplot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/plplot-devel
