You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Certain kinds of font testfiles are more easily and more accurately generated programmatically than by hand.
Examples (from Arabic projects) include: showing every character in the font in all relevant contextual forms (to test shaping logic) along with all combinations of features that affect any particular character (to test feature logic); showing representative diacritics on all possible bases and then all diacritics on representative bases (to test APs); etc.
Programs that generate such testfiles are dependent on a number of sources of data including the font(s) being tested, project-specific info about what characters are supposed to be in the font and what characters affected by what features, Unicode character properties (including those in the pipeline), etc.
Some of the dependencies are themselves products of the build process, and of course testfiles need to be regenerated when their dependencies have changed. An example from the wscript of the LateefGR project:
While this works, the .FTML files are generated in the results folder. Currently, for smith to use such generated .FTML files for a test target, they have to be moved to tests folder, but that is really a violation of the separation of source and results.
One possible approach is for smith to put/copy all testfiles (including both static testfiles from source and dynamically generated testfiles) into a 'results/tests' folder.
No description provided.
The text was updated successfully, but these errors were encountered: