There was a regression with 45c03413 (hal_json_style: prevent a
translation of worklist with document count, 2021-05-10), the title
of the worklist was translated twice. This was supposed to be
catched by the test, by checking that the mocked gettext had only
one call for "Draft to Validate", assuming that the translation of
"Draft to Validate" was "Draft to Validate" and that a double
translation would cause this to be translated twice, but because of
a side effect of another test inserting a translation for
"Draft to Validate", this problem was not noticed.

This fixes the double translation and adjust the test to not depend
on the actual content of message catalog.
Co-authored-by: Georgios Dagkakis <georgios.dagkakis@nexedi.com>

Grouping feature checks that the sum of all selected lines == 0, which
is often not the case as the values are float. For that, our approach
is to round the values with the precision of the accounting currency,
since these precisions are usually small (typically 0, 2 or 3), we
don't have problems with rounding. Using the section currency is not
just a workaround for rounding, it's also correct because we don't
consider more precise amounts in accounting transaction lines.

The problem with this approach was for the case where no accounting
currency is set on the section organisation, in that case we did not
round and this sometimes led to "grouping is impossible" errors that
are hard to find for users. At this level it's better to use a default
rounding precision that would make it possible to use the grouping
feature even when section currency is not set.

Some objects, in our case BTrees.Length.Length in a ZODB connection
have a __str__ method that returns unicode on python2:

u'<BTrees.Length.Length object at 0x7f850932e0d0 oid 0x1e334 in <Connection at 7f854bc0f190>>'

They cause an unicode error in the history view when they are
concatenated together with other str (encoded as UTF-8) properties,
this can be observed when using history view with a "folderish"
document (but not with a File as in test_ZODBHistoryBinaryData).

To prevent this issue, we use the fact that ''.format unlike '' %
seem to apply a str() on arguments and use it instead.
Co-authored-by: Yusei Tahara <yusei@nexedi.com>

Some objects become unicode when %s is applied and may cause
UnicodeDecodeError, thus use %r insead.

This reverts commit 40fded3e.

It is not good to mix unicode and str for UI. Use utf8 str only.

This should detect any Business Template with non-migrated workflows.
Exclude the Business Template which is precisely made to test workflow
migration.

Migrate oauth2_session_refresh_interaction_workflow to new-style workflows.
This time, the migration should be complete.

This reverts commit e925dc0b.
This commit is breaking more than it is fixing. Revert until I have a
working version of this migration.

Migrate oauth2_session_refresh_interaction_workflow to new-style workflows.

display comment by default

check contributor's organisation ealier

display corporate name instead of title

display web site instead of default document

add logo field for leaflet

display value used in dialog

update css

ValidationFailed is now a "real" class from Products.ERP5Type.Core.Workflow

See merge request nexedi/erp5!1676

this only supported http and never supported https, probably nobody is
using this and it makes porting to python3 more complicated

we had a patch to close <input/> elements, but with HTML5 this is no
longer recommended.

template_keep_path_list, template_keep_workflow_path_list and
template_keep_last_workflow_history_only_path_list were not sorted
for no reason

The _getOrderedList approach was not ideal, because the properties were
saved when being displayed, so the typical workflow was:

 1. enter the property as non sorted
 2. click save (property is saved as non sorted)
 3. the page displays the property again as sorted
 4. click save again so that the property is saved as sorted

by sorting the properties at save time, step 1 is enough

This change back the accessors to be "standard" accessors, ie. returning
lists, like every other list accessors, so a few tests and a few scripts
had to be adjusted for the new API.

Lazy loading img are also not loaded, which speed up page rendering.

This should evolve to become a browser-side OAuth2 client.

So every new instance is able to use self-contained oauh2 authentication.
In turn, this triggers atomated migration of a few portal types, which
cause the coding style tests to fail. So commit these as well.

Replace CookieCrumbler's cookie with OAuth2 tokens:
- add the notion of authenticated sessions: period from a login action to
  either a logout or the expiration of tokens
- avoids session fixation (knowing one token does not grant near-permanent
  access to the session)
- reduces the per-request cost of authentication (removes the need to
  compute user's groups, and the need to cache these groups for a fixed
  period).

Also, to reliably open login_form, log out.

This is a simplified version that works well enough for python 2.
For Zope 4 and python 3 we are considering reusing publish.