pmezard at gmail.com
Wed Dec 26 15:47:40 CST 2007
Is there a canonical way to write extensions tests and run them ?
For instance, the forest extension module and test-forest are in the same directory (there are also other tests in tests/). Calling run-tests.py from the extension root directory works fine but has several limitations:
- tests and tools living in the same directory is no longer true. It means extension tests cannot use mercurial tools like hghave or simplemerge. Test coverage cannot be generated.
- the extension code is not installed. It works with forest because run-tests.py is executed in the extension module directory, which is added to sys.path. I am not convinced requiring the extension code to be with tests is a good practice.
- what about extensions with a setup/compilation process ?
First point can be solved with a separate environment variable, but I prefer to hear extension writers before moving forward. It would be great that a collection of extensions could be tested against a given mercurial version without installing any of them.
More information about the Mercurial-devel