On 2013-12-16 22:58-0800 Alan W. Irwin wrote: > IMPORTANT: Also, I have made so many build-system changes since you > guys last tested, this is advance notice that later this week after > all my tests are done (including the currently non-working or > non-tested epa_build environments for both Linux and MinGW/MSYS), I > will call on as many of you as have time for it this close to the > Christmas season to do one last round of testing on your own > platforms/build system environments before I release. > > The reason this additional testing is important is my recent results > show that comprehensive tests can sniff out build-system bugs much > better if applied for a wide variety of build environments i.e., cmake > options, environment variables, buildtools that have been built, > dependencies that have been built, and system libraries that are > available.
My three build environments are (1) My normal one which uses Debian Wheezy installed packages for essentially all PLplot dependencies. So for this case, PLplot is full-featured but the buildtools and dependencies tend to be old versions. (2) The "build_plplot_lite" epa_build environment which uses epa_build buildtools results such as Tcl/Tk8.6, epa_build results for many ordinary but minor PLplot dependencies (like shapelib and qhull), but which specifically ignores cairo, qt, and wxwidgets. So for this case, PLplot is not full-featured. (3) The "build_plplot" epa_build environment which uses epa_build buildtools results such as Tcl/Tk8.6 and epa_build results for essentially all ordinary PLplot dependencies (including all the minor ones such as shapelib and qhull, and all major ones, i.e., Qt4, the pango/cairo libraries, and wxwidgets, except for octave. So for this case PLplot is full-featured like in (1), but the dependencies are newer versions that have been build by epa_build rather than system versions. On Linux I have now successfully completed default (where essentially all possible tests are run for the particular build environment) runs of scripts/comprehensive_test.sh both for (1) and (2). However, (3) is currently not working because of a rather nasty dependency mess when attempting to epa_build both pango/cairo and wxwidgets on Unix. The issue is that wxwidgets normally uses the gtk toolkit on Unix so wxwidgets should be built and linked against a consistent version of gtk. But epa_build currently does not supply that (just a consistent pango/cairo subset of gtk). So the resulting epa_build of wxwidgets was linked against some system libraries that are part of the gtk suite of libraries and the epa_build pango/cairo subset of the gtk suite of libraries, and that mixing of inconsistent library versions was causing test errors for wxwidgets in the case of (3) above. I explored the possibilities of using the bare X toolkit and also the motif/lesstif toolkits instead (to force using pure system libraries for the epa_build of wxwidgets), but those wxwidgets epa_builds failed because those toolkits are not unicode aware and PLplot needs a unicode-aware version of wxwidgets. So the only wxwidgets toolkit that we can use on Unix is gtk, and the long-term solution for that case is to epa_build all of the gtk suite of libraries in a consistent way. In principle, that should be completely straightforward since I have all the build data and dependencies for those libraries at my fingertips, but in practice I will wait to implement that until post-release since it will probably take several days to refine those build data so that the many different libraries in the gtk suite all epa_build properly on Linux. In the short term for this release I am going to shut down the epa_build of wxwidgets, and then (3) will end up using the (consistent) system version of wxwidgets, and the wxwidgets-related errors that were stopping (3) from being a success _should_ disappear. Assuming I can get that to work early tomorrow (Wednesday), then my last pre-release step is to get (2) to work on MinGW/MSYS/Wine but the timing of finishing that is still quite uncertain (it could be Wednesday night or several days later) because it has been so long since I have tried the MinGW/MSYS/Wine case. However, I will continue to keep you informed on the timing of this last pre-release work as that timing gradually becomes more certain. Alan __________________________ Alan W. Irwin Astronomical research affiliation with Department of Physics and Astronomy, University of Victoria (astrowww.phys.uvic.ca). Programming affiliations with the FreeEOS equation-of-state implementation for stellar interiors (freeeos.sf.net); the Time Ephemerides project (timeephem.sf.net); PLplot scientific plotting software package (plplot.sf.net); the libLASi project (unifont.org/lasi); the Loads of Linux Links project (loll.sf.net); and the Linux Brochure Project (lbproject.sf.net). __________________________ Linux-powered Science __________________________ ------------------------------------------------------------------------------ Rapidly troubleshoot problems before they affect your business. Most IT organizations don't have a clear picture of how application performance affects their revenue. With AppDynamics, you get 100% visibility into your Java,.NET, & PHP application. Start your 15-day FREE TRIAL of AppDynamics Pro! http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk _______________________________________________ Plplot-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/plplot-devel
