On Thu, 2013-10-31 at 09:09 +0100, Éloi Rivard wrote: > Take your time :) > > If you can explain me how you would test it, maybe I can set up a > testcase.
Once we have passed the first test which is to generate an executable for the target I think we could potentially have two further types of test. The first would be to launch Denemo with scripts that would generate a LilyPond file and check for regression, a MIDI file and check for regression, and (some work would be needed for this) run a selection of the Denemo commands checking that the scheme scripts did not barf. The second would be to generate a representative score in pdf and display it (and the Denemo display) to the tester (who must be a person in this case). The "representative score" should be one that covers what we think are the most popular features, but which is sufficiently simple that looking over it has a reasonable chance of triggering error recognition. This would mean that it should make sense as music notation at least. This would still leave swathes of stuff untested: following links to open scores, creating new commands, changing shortcuts (even, do short cuts work at all), but we have to make a start somewhere and the above is quite easily do-able. I have made a start by fixing the non-interactive launch of Denemo for a couple of bugs, but even so I am not at all sure it would launch if there was no display for GTK to interact with, so these tests would need a person to launch them manually (well, you know what I mean)... Richard _______________________________________________ Denemo-devel mailing list [email protected] https://lists.gnu.org/mailman/listinfo/denemo-devel
