[Paraview-developers] Proposal for testing Lookmarks
Berk Geveci
berk.geveci at gmail.com
Thu Aug 18 08:07:57 EDT 2005
Hi Eric,
I am against doing this to all tests. Here are my reasons:
* ParaView tests take too long already. Because of this, nobody,
including me, runs all the tests before commiting. The best people do
is to run some of them. It takes over 1 hour to run all the tests on
my machine.
* Some of the features of the lookmarks are tests by trace tests. The
trace is already tested on quite a few tests. Adding lookmarks to all
tests is not going to increate lookmarks coverage by too much.
* Lookmarks, although a very very nice feature, is not a core feature
of ParaView and should not take so much of ParaView testing time.
* It is best to create tests to increase the coverage of
vtkPVLookmarkManager. This class has the largest number of uncovered
lines in all ParaView (1570 !). Adding these lines to all tests is not
going to increase coverage of this class. I would suggest writing one
or two specialized tests to increase coverage. These will probably
require driving the GUI from the script.
* There is no point in testing lookmarks in all the parallel tests
since none of the lookmark features are parallel specific. The
parallel dashboards spend 2/4 to 3/4 of their time running parallel
tests.
I would suggest picking a few (4-5) tests from ParaView and adding
this to them. The ones that had problems to start with are good
candidates. Also, creating 1 or 2 tests that increase the coverage of
vtkPVLookmarkManager would be very useful.
-Berk
On 8/17/05, Stanton, Eric T <etstant at sandia.gov> wrote:
> Greetings,
>
> As a step towards improving the quality of the lookmark code, I'd like
> to get the go-ahead to add the following lookmark commands to every
> paraview tcl test script in the .../GUI/Testing/Tcl directory just
> before the image comparison takes place.
>
> # Lookmark Test
> set lookmarkManager [[ $Application GetMainWindow ]
> GetPVLookmarkManager]
> set lookmark [$lookmarkManager CreateLookmark name-of-test 0]
> [ $Application GetMainWindow ] DeleteAllSources
> $lookmark View
> $lookmarkManager Withdraw
>
> I have run the tests with these commands added and have identified and
> fixed several bugs in lookmarks involving incompatibility with certain
> kinds of readers, incorrect coloring of actors, etc. The built in image
> comparison is ideal for checking that a lookmark has worked correctly
> since the idea of a lookmark is to take you back to the exact view you
> left off at.
>
> There are still a handful of tests for which creating and viewing a
> lookmark fails (those involving animation, group input widgets) and some
> for which they don't make sense. I will only commit tests that have
> succeeded on the machines I've tested and continue to debug the others.
>
> I have written a couple tcl scripts that do the adding and removing of
> lookmark commands. They can be run when a new test is added to the
> directory.
>
> I am also working on another testing option in which the lookmark test
> script will open unmodified test scripts, add the lookmark commands to a
> temp file, then run the modified temp script. There are some issues with
> this however that are still being worked out.
>
> The downside to all this is that it increases the testing time. But I
> think the benefit of improving the quality of lookmark code outweighs
> the costs.
>
> I am also open to any suggestions anyone might have.
>
> Regards,
>
> Eric Stanton
> Sandia National Laboratories
> Data Analysis and Visualization
> 505-284-4422
>
> _______________________________________________
> Paraview-developers mailing list
> Paraview-developers at paraview.org
> http://public.kitware.com/mailman/listinfo/paraview-developers
>
More information about the Paraview-developers
mailing list