On Fri, Mar 05, 2010 at 01:52:50PM -0800, Alan Irwin wrote: > The "p" octave examples were designed many years ago by Joao Cardoso (since > retired from PLplot). At that time, -dev xwin worked for most/all of those > "p" examples and gave some nice-looking interactive effects. (Joao had an > extraordinary artistic ability to make good-looking examples.) > However, -dev psc did not react well to some of those interactive effects so > in our non-interactive tests that use -dev psc we have automatically dropped > many of those "p" examples from consideration. > > I have recently implemented test_octave_$device targets for all our > interactive devices in an attempt to test _all_ the "p" octave examples for > our interactive devices. It uses the configured script plplot-test.sh(.in) > with option --interactive_octave to make a call to the configured script > test_octave_interactive.sh(.in). It's that latter script that must be > modified if you want to change the list of "p" examples that are tested by > test_octave_qtwidget. (I temporarily dropped some of those examples from > consideration because of various issues that I commented in > test_octave_interactive.sh(.in).
Alan, This is a really useful addition - thanks. > I mention that interactive test target specificially because -dev qtwidget > currently is the best of all our interactive devices at interacting with > octave. For example, some kind of bit rot has set in and test_octave_xwin > and test_octave_tk (which exercise -dev xwin and -dev tk for these > interactive octave tests) both error out immediately. That is a shame > because I remember that it used to work (reasonably) well (at least for -dev > xwin, but I am sure I tried -dev tk back then as well). test_octave_xcairo > is painfully slow (this interactive issue for -dev xcairo has been reported > independently as bug 2960406) on my platform so I didn't have the patience > to try the test_octave_xcairo target. test_octave_wxwidgets works fine on > individual examples, but currently there is some incompatibility between how > the p examples are exited and reentered in octave and -dev wxwidgets so for > any list of examples that has more than one example, test_octave_wxwidgets > always segfaults on the second example tried. The xwin problem is simple - and nothing to do with the p?? examples. Your script called figure with a file name set. This doesn't make sense for interactive drivers. Worse, the xwin driver interprets this as a display to try to open. This obviously fails - hence the xwin problems. (As an aside, this is odd behaviour and I don't like it, but it has been there for a long time. We should be consist with other X-based drivers at least.) > > I have reported these general device driver/octave issues as well as a > number of individual issues for individual "p" examples that have been > turned up by this new test on our bugtracker (bug 2964497). I have assigned > this bug to Andrew because he is our octave go-to guy, but I encourage > anyone else here with an interest in octave (or the xcairo, xwin, and > wxwidgets devices) to also tackle this list of issues. > > The first goal should be to extend the range of "p" examples working for > qt_widget. There is some low-hanging fruit there where the plot GUI comes > up, goes through its interesting interactivity, but then hangs on exit > (presumably because the "p" example in question doesn't exit properly). Once > p1 through p21 are working for qtwidget, then it would be good to get xwin > working properly with these octave examples again and then the remainder of > our interactive devices. Also, I am sure as more interactive devices are > made to work with these examples some design decisions that Joao made (sizes > of legends, for example) to tune the examples to work with -dev xwin will > have to be reviewed to give a polished look for our other interactive > devices. Ultimately, we may decide to reinstate the full list of "p" > examples that are exercised in the standard noninteractive test as well. I think a lot of the other issues are to do with the interactive aspects. Only the xwin and tk drivers support xor mode which is required. I could work round this (as example 20 does), but this hides the issue. It would be really nice to add this capability to qtwidget and xcairo. There are still other issues with the octave scripts which I will continue to investigate. Andrew ------------------------------------------------------------------------------ _______________________________________________ Plplot-devel mailing list Plplot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/plplot-devel
