On Thu, 2014-05-15 at 17:51 +0200, Éloi Rivard wrote: > > > > 2014-05-14 18:43 GMT+02:00 Richard Shann <[email protected]>: > On Wed, 2014-05-14 at 18:15 +0200, Éloi Rivard wrote: > > > > The reference file should be in integration-data. > > > > Comparison against reference files is done for .denemo files > in > > example directory, and scm / mxml files in integration-data > > > > Sorry but I still don't understand. The file hemiola.denemo > which I have > checked into tests/integration-data is not suitable for > inclusion in the > example directory - > > > I see, I will look for a way to fix that.
No need! It is "idempotent" so it is fine in the the integration-data directory. (I hadn't read your Readme.md file when I typed that). I have added another file in tests/integration-data that does some general note insertion, and that again has a script inside it which re-creates itself. The build passed so I assume it got loaded saved and the result was the same as the original file, which means that a whole set of commands for inserting notes (going over to the next bar as needed) and several other things are now tested for regression. What we *could* have is files which alter themselves, but I think you have created something perhaps better: we can have a scheme script which generate Denemo file. That is suppose I create mytest.scm containing (d-InsertA) (d-Save "filename=mytest") Then and put it in integration-data along with mytest.denemo that results from the script. Will this then pass? Shall I try checking it in to see? Richard > > > > > it is a test, not something users will want to look > at. It is, in fact, the same as its output, that is if you > load it and > then save it then the saved file is the same as the original, > or should > be if the command it tests is still working. > Ah! When in doubt, read the documentation :) : > From the Readme.md file this single file should be good as a > complete > test. It would not be good to have a file which when loaded > and saved > became different, such a test would need to go elsewhere and > would > presumably require more infrastructure. And this is not really > needed, > since such a test could always be done via a scheme script. > > Ok, so if I have understood correctly, the check in I did of > hemiola.denemo was all that was needed, and the travis-ci > report that > the build passed means that the test passed. > Excellent! > Richard > > > > > > > -- > Éloi Rivard - [email protected] > > « On perd plus à être indécis qu'à se tromper. » > _______________________________________________ Denemo-devel mailing list [email protected] https://lists.gnu.org/mailman/listinfo/denemo-devel
