On Tue, Dec 16, 2008 at 09:59:25PM -0800, Alan Irwin wrote: > Hi Andrew: > > Would you please have a look at the results for a substantially modified > 31st example that I put together in python? > > The principal changes from your implementation in C are as follows: > > (1) I took all your pre-plinit tests and spanned them between pre- and > post-plinit to check if the device driver molested any of the values that > were specified. > > (2) All "get" values are output to stdout before each test so it is easy to > see when things go wrong and see what tests have been run. > > (3) I massaged some of the set values to be more unique (i.e., to > make sure there was no confusion between x and y values, and to avoid > zero values and unity values). > > Immediately interesting results are two of the tests failed because of some > screw-ups in the python interface. So exposing those interface bugs > immediately justifies this effort. However, the fix for these issues was not > at all obvious to me. Could you have a look to see whether you can spot what > is wrong? > > Since I could not make quick fixes, I commented out the plexit() and > continued with further tests in each of these cases. Once we sort out what > is wrong with the python interface, then those commented commands will have > to be uncommented. > > One of my motivations for (2) above was to insure the same tests were run > for all languages since currently we have no knowledge of that except by > reading through the various example 31 implementations. However, a > comparison between stdout results (presumably for examples 1 and 23 as well > which also output to stdout) would need some extra machinery in the compare > ctest. Would you be willing to implement that? > > If you (or anyone else) wants to do any more changes to how the existing > tests are done in xw31.py or add some additional tests to that example, > please feel free. Once we all agree that xw31.py is perfect, I still think > we should delay propagating it to the other languages until a day or so has > gone by to make sure we don't have any more bright ideas.
Alan, I like what you have done. I'll look at comparing the stdout output as part of ctest. A couple of other examples also produce this, so it would be worth checking. There will be slight complications with the -V option and with the output from plgfnam, but I think we can work round that with judicious use of stderr for this. It's a slight fudge, but The stdout test would also be complicated if you used the -V option or printed out the file name from plgfnam, since these will differ between languages. One further thing - I would save / restore the familying settings well as the page settings otherwise you end up with file names like x31p10.psc rather than x31p.psc in the plplot_test directory. This messes up the compare tests. Actually, plsfam must be called before plinit so we can't easily test this across plinit. I suppose we could test turning off familying. We only produce 1 blank page so we don't need it and this would ensure the postscript file names were ok for ctest to check. I've implemented this in the python version. Andrew ------------------------------------------------------------------------------ SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ _______________________________________________ Plplot-devel mailing list Plplot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/plplot-devel
