:date: 2017-10-04

==========================
Wednesday, October 4, 2017
==========================

Avanti advances
===============

I continued to work on :ticket:`2083`.

New rule for the :class:`EnrolmentChecker
<lino_avanti.lib.courses.EnrolmentChecker>`: when a reminder exists,
absences before that date are ignored.

The workflow for reminders is currently just my first quick draft.

New template files
:xfile:`courses/Enrolment/Default.odt`
and
:xfile:`courses/Reminder/Default.odt`
       
Side effect: The default :meth:`get_slave_summary
<lino.core.actors.Actor.get_slave_summary>` now adds an insert button
(if an insert_layout exists and user has permission).

DummyField now has a :meth:`__str__` method.

New fixture demo2 which adds primary coaches to the clients one
reminder.

I then did a release during their working time (using the fact that
they had team meeting) and did local configuration of the template
files. Unfortunately Johanna and I failed to meet afterwards because
Hangouts wasn't in the mood today.

ExtJS 6 support advancing
=========================

There are still many details to fix before we can consider ExtJS 6 to
be done... but we are advancing. Tonis and I fixed a few things
together.


Making :ref:`appy` ready for Python 3
=====================================

I had a nice shared screen session with Gaetan, the autor of
:ref:`appy`.  Together we managed to make appy pass the
`test_appy_pod` test of the :mod:`lino_xl` suite.  Our work is published on
forge.pallavi.be: revisions `62
<https://forge.pallavi.be/projects/appy-python-3/repository/revisions/62/diff/>`__,
`63
<https://forge.pallavi.be/projects/appy-python-3/repository/revisions/63/diff/>`__
and `64
<https://forge.pallavi.be/projects/appy-python-3/repository/revisions/64/diff/>`__.

This is great news because :ref:`appy` is still the best and only real
solution for generating .pdf files from .odt fies.

The problem with Gaetan is that he uses a rather unorthodox packaging
system.  I knew from earlier discussions that trying to convince him
to become normal is just a waste of time.  So I didn't do that, and we
found a solution for every problem.

The *clou* of his strange system is that he refuses to have a
:xfile:`setup.py` file in the root directory of his project. Don't ask
me why.  To install a development version of appy, you must add a
symbolic link to your :xfile:`site-packages`.  This causes at least
two problems:

- the `setup.py test` command tries to install appy from PyPI because
  it doesn't know that it actually is installed. We fixed this by
  temporarily removing ``'appy'`` from the :term:`install_requires`.

- we cannot specify the development snapshot as requirement on
  :ref:`travis` because Travis cannot install without a
  :xfile:`setup.py` file. This means that Gaetan must
  release our work on PyPI before the :ref:`travis` tests will
  pass. But okay, we are not about a few days now that we fixed
  a problem that
  has been waiting at least a year.


Gaetan reminded me the basic operations of Subversion which I have
been using much but hadn't used for years::
  
    $ svn up                # git pull
    $ svn ci -m "message"   # git ci -m & git push

And he also reminded me a debugging technique which I hadn't used for
years now::

    import pdb ; pdb.set_trace()

And then in the debugger you need to know at least four commands:

- l : list
- n : go to next line
- s : step into
- c : continue

Now that the :ref:`xl` test suite passes (at least on my machine), I
continued with the test suite of :ref:`book`. This one still has
dozens of failures which are *not* caused by appy.


Release to PyPI
===============

I released some projects to PyPI just because I the information there
was *so* obsolete.  Allthough actually this makes not much sence, we
don't currently recommend using a released version of Lino.

OMG, and now the information there is completely ugly:
https://pypi.python.org/pypi/lino