The Mystery of the failing Welfare test suite (continued)

Saturday, April 10, 2021. Continued from Friday, April 9, 2021.

This morning I discovered that my observations yesterday about wrong ordering came because I had a bug in my new version of the lino_welfare.modlib.welfare.models.customize_sqlite() function: it mixed up -1 and 1. That’s the explanation for the mysterious observations I saw yesterday.

I wrote a new doctest docs/specs/topics/sorting.rst (rendered About sorting) in order to verify my theory that the issues are caused by demo data being generated differently because sorting behaviour has changed. And the result is negative, i.e. my theory seems wrong. This file passes always, even with the pypi versions of lino, xl, welfare and weleup.

Right now I have no idea what else might be the reason.

When run with Django 3.1.8 and Lino master, inv prep test in welfare yields 2 failures, one being caused because django.db.models.fields.AutoField changed to django.db.models.fields.BigAutoField, the other is the total number of cal.Event objects, where the docs/specs/weleup/eupen.rst expects 311 but gets 588. This failure might have been there for some time, there is indeed

I now added show_db_overview() to docs/specs/weleup/eupen.rst.