On 2015-02-15 16:54-0000 Arjen Markus wrote: > The same [degenerate path] pattern appears (for the Rosenbrock function) in > the Tcl version, but I have no idea what it is supposed to draw and why the > Fortran version does not show this. All the calls are there in all versions. > Really odd. If anyone has fresh ideas where to look for the discrepancy, I'd > be obliged.
To summarize my previous response to you about example 8 consistency issues: (1) the test_noninteractive target and all other parts of the test suite consistently use the Sombrero function so you should ignore Rosenbrock results from now on, and (2) if the test suite says you are getting inconsistent (Sombrero) results for any language, then likely a small adjustment of the cutoff ellipse size (consistently done for all langues) in page 5 and 10 of the example will fix the issue. >> >> Thanks very much for your willingness to do those comprehensive tests, and I >> look >> forward to seeing your results for those tests. >> > > There is something extremely weird going on with the make utility that comes with MinGW, but the workaround is to use "mingw32-make" instead of "make". That is not a workaround or anything weird. MinGW without MSYS (the "MinGW Makefiles" generator) is very different than MinGW with MSYS (the "MSYS Makefiles" generator). For the former you must use "mingw32-make" (which appears in the MinGW PATH). For the latter you must use "make" (which appears in the MSYS PATH). > Cygwin gives nice clean results using "make test_noninteractive", except for > example 8 (see above). I am pretty sure a small adjustment of the cutoff ellipse size (consistently for all languages) will solve that remaining issue. > Under MinGW I do not have the test scripts, which is odd (I have not > investigated this point yet). Not odd. The test scripts require many unix tools such as bash. So they are only deployed if bash, etc., are available on your PATH. So put MSYS (which includes bash.exe and all the rest) on your PATH, use the "MSYS Makefiles" generator, and you should be fine. If you use a hack I discovered, you can also use the "MinGW Makefiles" and the test suite with most of but not all of MSYS on your path, but let's not go until post release and instead concentrate now on getting good test results for both MinGW/MSYS with the "MSYS Makefiles" generator and Cygwin with the "Unix Makefiles" generator. > All the testing indicates that apart from the oddities in example 8 everything is working fine. I will check with the sombrero function. Good. Sombrero is all you should look at. > Another curiosity about MinGW: it seems the symbols are exported automatically, just as in Cygwin and Linux. No need for the "definitions file" (.def) anymore. This seems a change from previous versions. Interesting indeed. Of course, you will want to keep the MinGW .def file up to date for those users (like me) who are using an older MinGW version. In conclusion, I think the above test results from you are quite promising, and there is a good chance with ellipse size adjustments of getting absolutely clean test_noninteractive results for both MinGW/MSYS (with the "MSYS Makefiles" generator) and Cygwin. That's a huge step forward, and you might want to rest on your laurels there. However, once you have a clean test_noninteractive result for example 8, then I hope you would also be willing (for both the MinGW/MSYS and Cygwin platforms) to run scripts/comprehensive_test.sh as well. That script by default runs ctest, the test_interactive target, and the test_noninteractive target for our build tree and two versions of our installed examples tree for our three major configurations (shared libraries/dynamic devices, shared libraries/static devices, and static libraries/static devices). Use the --help option for that script to learn more. Once previously in this release cycle I have run scripts/comprehensive_test.sh in default mode with complete success for MinGW/MSYS on the Wine version of Windows. But I hope you are willing to do it because it took several days on my slow Wine version of Windows compared to only an hour or so on Linux, and it should only be a couple of hours for you (assuming Microsoft Windows is roughly the same speed as Linux). 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 __________________________ ------------------------------------------------------------------------------ Dive into the World of Parallel Programming. The Go Parallel Website, sponsored by Intel and developed in partnership with Slashdot Media, is your hub for all things parallel software development, from weekly thought leadership blogs to news, videos, case studies, tutorials and more. Take a look and join the conversation now. http://goparallel.sourceforge.net/ _______________________________________________ Plplot-devel mailing list Plplot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/plplot-devel
