Tuesday, September 12, 2017

Two problems after release@cpaschatelet

I continue to like the idea (Monday, September 4, 2017) of having no phantom row in tables that have a (+) button (IOW an insert_layout), but there is at least one place where it disturbs: when users used a double-click to that the phantom row from within a slave table. New attribute no_phantom_row. But I realized that the only real solution to Lydia’s problem would be a new argument in the AJAX call where the grid component tells Lino whether it is a main component (with toolbar) or a slave panel in a detail window (without toolbar).

cal.Events was using two variables pending_states and fixed_states when observed_event was EventEvents.stable EventEvents.panding. But these lists of states were being filled (using EntryStates.filter(fixed=False)) at module scope, i.e. when models were getting imported. That might have caused some troubles for them: J’ai trouvé une anomalie pour le filtrage des rendez-vous quand un “Évènement observé:” est spécifie.

Adapting test suites

I released commondata.ee version 0.0.4 because version 0.0.1 failed on Python 3. Versions 0.0.2 and 0.0.3 were never released to PyPI due to misunderstandings with the Trove classifiers (“BSD License” must now have a capital L and natual language estonian doesn’t exist).

I removed the is_local_project_dir Site attribute because it is no longer needed. The local config dir is always a subdirectory config of the site_dir

The solution for #1296

Finally I found the solution for #1296! The problem occurred during the Python3 subprocess spawned by appypod when converting the generated odt file to pdf. The problem occurred only when running our test using setup.py test.

It was because somehow the PYTHONPATH of the calling Python 2 process was inherited to the Python 3 process given by pythonWithUnoPath. That caused the Python 3 process to import modules of Python 2. And especially future doesn’t like being imported by a Python that didn’t install it). I have no proven explanation, but my guess is that setup.py test sets the PYTHONPATH.

atelier.test.DocTestCase() now explicitly removes PYTHONPATH from the environment of the child process.

Kontoauszüge erfassen

  • ‘DueMovement’ object has no attribute ‘obj2href’

  • quick search fields for financial voucher items

  • Activate Merge action