20130421 (Sunday, 21 April 2013)

“Upcoming” and “Missed” events

The coming_reminders and missed_reminders virtual fields of lino.modlib.cal.models.Home (“Ausblick” und “Verpasste Termine”) showed also events of other users.

Added a new case to Miscellaneous and fixed the bug.

Which revealed that this bug is probably at least one release old. The fact that nobody complained indicates that these two panels are obsolete and should be replaced by something better.

Sphinx: unsupported build info format in .buildinfo

I sometimes had the following Sphinx warnings:

WARNING: unsupported build info format in u'/home/luc/hgwork/welfare/docs/.build/.buildinfo', building all

Which disturbed me because it stopped the build (because I have -W option turned on and want it to stay like this).

This invalid .buildinfo file contained:

# Sphinx build info version 1
# This file hashes the configuration used when building these files. When it is not found, a full rebuild will be done.
config:
tags:

It is indeed not normal to have this file exist with “empty” entries. Possible that this comes because I use complex configurations. But still it was Sphinx herself who generated this file and now she complains about the content. That’s not correct! Here is the responsible code in /sphinx/builders/html.py:

try:
    fp = open(path.join(self.outdir, '.buildinfo'))
    try:
        version = fp.readline()
        if version.rstrip() != '# Sphinx build info version 1':
            raise ValueError
        fp.readline()  # skip commentary
        cfg, old_config_hash = fp.readline().strip().split(': ')
        if cfg != 'config':
            raise ValueError
        tag, old_tags_hash = fp.readline().strip().split(': ')
        if tag != 'tags':
            raise ValueError
    finally:
        fp.close()
except ValueError:
    self.warn('unsupported build info format in %r, building all' %
              path.join(self.outdir, '.buildinfo'))
except Exception:
    pass

A principal problem with this code is that it uses the standard exception ValueError to handle a custom problem. It’s a mousetrap, and my concrete case of invalid build file makes Sphinx step into this trap. In fact (AFAICS) the author wants Sphinx to warn only in those special cases and to silently ignore (without any warning) any “really invalid” .buildinfo file. So I suggest to replace “ValueError” by a custom exception “InvalidInfo”:

class InvalidInfo(Exception):
    pass

try:
    fp = open(path.join(self.outdir, '.buildinfo'))
    try:
        version = fp.readline()
        if version.rstrip() != '# Sphinx build info version 1':
            raise InvalidInfo
        fp.readline()  # skip commentary
        cfg, old_config_hash = fp.readline().strip().split(': ')
        if cfg != 'config':
            raise InvalidInfo
        tag, old_tags_hash = fp.readline().strip().split(': ')
        if tag != 'tags':
            raise InvalidInfo
    finally:
        fp.close()
except InvalidInfo:
    self.warn('unsupported build info format in %r, building all' %
              path.join(self.outdir, '.buildinfo'))
except Exception:
    pass

More bugs fixed

  • pcsw.Client.print_eid_content didn’t work.
  • Added a test case to General PCSW.

Merging imported Partners

Es war ein Denkfehler, die Aktion “Fusionieren” für importierte Partner zu verbieten. Zumindest im Fall “23219 nach 23624 fusionieren” ist dieses Verbot falsch. Also raus mit folgendem Code (aus der lino_welfare.modlib.pcsw.models.Partner):

def get_row_permission(self,ar,state,ba):
    if isinstance(ba.action,dd.MergeAction) and settings.SITE.is_imported_partner(self):
        return False
    return super(Partner,self).get_row_permission(ar,state,ba)

Oho, another surprise: adding MergeAction in post_analyze is too late, but adding it in pre_analyze is too early! See also 2013-04-09: MergeAction needs the info in _lino_ddh to fill keep_volatiles. Solution: pre_analyze is now being emitted a little bit later: after setup_choicelists(), setup_workflows() and the loop which fills _lino_ddh.

TODO:

  • File does not exist: /usr/local/django/cpas_eupen/media/eid-jslib/media
  • Shouldn’t we remove allow_cascaded_delete = [‘client’] on pcsw.Coaching? because it sounds more intuitive that you cannot delete a Client without first manually deleting every Coaching.