In case of ConflictError, request will be retied for
HTTPRequest.retry_max_count times. For Zope 2, the default value of the
class attribute was 3, but on Zope 4, default value is 0 and this is set
to 3 on startup (by Zope2.Startup.handlers.root_wsgi_handler).
This startup code is not executed by runUnitTest, we need to set
HTTPRequest.retry_max_count somewhere, otherwise ConflictError are not
retried during tests.

The status message no longer use system-msg class, use simpler
"assertTextPresent" that should be enough.

Simplify the waiting for static pages, clickAndWait is enough.

I'll make a commit to handle all tests

When rendering a proxy field, 3 different fields can come in play:
1. the field to be rendered
3. the template field (i.e. not a proxy field) that knows how to render
2. possibly an intermediate proxy field that contains the value to render

What's difficult when rendering a proxy field is to take the above 3 fields
into account and this commit does it by creating a temporary field:
1. 'field' variable in TALES
2. the value
3. the code

Before this commit, 1 could be wrong.

This was deprectated because we don't have get*ById for other modules
and tools, we just use OFS API. This should also be slightly faster
because one less method call (and one less call to warning)

exiting the context manager restores the original DateTime behavior

This tests the test framework, because this code is tricky.

I'm not sure if this is used, but not being valid python code cause
problems with code analysis tools

mock has several advantages, the main one here is that it errors when
patching does not replace an existing attribute, which happens when we
are not patching the right place.

These attributes are not the same place in DateTime 2 and 3, this code
when using DateTime 2 was replacing the attributes, but in DateTime 3 it
was just creating new attributes that were never used. Update the code
to patch the DateTime 3 location

Maybe this made sense long time ago, but nowadays we are using
equivalence testers which tolerate date differences with more flexibility.

createDateTimeFromMillis was also problematic as it uses internal
private attributes of DateTime which is a pylint error with more recent
DateTime

Permissions are managed by the document_publication_worflow

Most service worker precache scripts reference a favicon.ico, but this
is using the default favicon.ico from Zope and even though it was included
in all ERP5JS and OfficeJS web sites, this was mostly not used, only
web_renderjs_ui web pages reference favicon.ico.

There's a favicon.ico in erp5_xhtml_style skin folder, but the skin
folder is not in ERP5JS skin selection.

On Zope2, this caused ERP5JS and OfficeJS application use the default
Zope favicon. On Zope4, the service worker can not fill its cache
because of 404 errors, because since Zope commit 4f0770941 (Retired
icons from the `Zope Management Interface` and various smaller cleanups
of ZMI screens., 2011-07-02) there's no default favicon.ico anymore.

To solve this, provide a favicon.ico in ERP5JS skin selection, by
copying the one from erp5_xhtml_style. We also reference it explicitly in
web site layout properties so that it remain in the cache.

OfficeJS applications do not use favicon.ico explicitly. They use icons
in their web application manifest, but this does not seem to be use as
favicon unless the PWA is installed. This part is not addressed by this
commit.

With zope.tal 3.6.0 - commit 26fadc4 (Increase the default value of
the `wrap` argument from 60 to 1023 characters, to avoid extra
whitespace and line breaks., 2011-08-20) from zopefoundation/zope.tal -
page templates keep the attributes on the same line and produce
slightly different output, although equivalent.

This commit updates the places where we do snapshot testing to the
new output. We discussed using lxml features to compare html and xml
from https://lxml.de/lxmlhtml.html#running-html-doctests but because
the corporate identity web pages are invalid html, it seems unsafe
to rely on lxml at this point. Once corporate identity web pages
are fixed this would probably be a good thing to do.

Avoid repeating processing_node condition for every single activity
whose serialization_tag dependency is being checked. This reduces the
length and execution complexity of produced SQL.

The meaning of empty items for **multi** list fields for categories is
not clear for dialogs (if user does not want to apply any filter, then
the natural way would be to select nothing).

This also caused issues with category fields, when the action script
uses restrictedTraverse to get the uids corresponding to the category
but a path is empty, like for example in 
https://lab.nexedi.com/nexedi/erp5/blob/d51bb0413a806b3db0c5eb69dec06065b9601322/bt5/erp5_accounting/SkinTemplateItem/portal_skins/erp5_accounting/AccountModule_getTrialBalanceReportSectionList.py#L40-48

which does this:

```python
# optional GAP filter
node_uid = []
gap_uid_list = []
for gap in request.get('gap_list', ()):
  gap_uid_list.append(portal.portal_categories.gap.restrictedTraverse(gap).getUid())
if gap_uid_list:
  node_uid = [x.uid for x in portal.portal_catalog(
                                   portal_type='Account',
                                   default_gap_uid=gap_uid_list)] or -1
```

If an empty item is selected, then `gap_uid_list` will contain an entry for 
`portal.portal_categories.gap.restrictedTraverse('').getUid()` which will be the
uid of the gap base category. Searching with a base category uid nowadays does not
match any document, but before 95e3eaec (CMFCategory: Do not index any Base Category
as a related document., 2016-12-21), it was matching all documents having a relation
from this base category and in the case of this trial balance report it was matching all
accounts.

This was a problem for old instances with accounts created before 95e3eaec, because when
they were first indexed, they had the record in category table, so they were matched, but
once they get re-indexed, they no longer had the record, so the result of this report when
selecting the empty item became different, because accounts were no longer included.

Looking back at this, maybe when updating to get 95e3eaec, we should have ran a migration
to delete all these records (re-indexing every document in the background should be enough)
so that if there's a problem, the problem happens right now and not after a few months
after accounts are modified and re-indexed.

When looking at this from end user level, theses empty items not only cause this problem,
but also does not have a clear behaviour and are not needed, so the changes here are about
removing these empty items.

In accounting reports, there was a multi listfield showing all gap categories, "grouped" by
chart of account - but the name of the chart of account was not displayed. This change to
use a None item, which is rendered as disabled to display the chart of account name, but to
do this we had to fix a bug in Formulator, these disabled items were only working properly
for single item widgets, not multiple items widgets.

See merge request nexedi/erp5!1572

This addresses a regression from nexedi/erp5!1561 changing the class hierarchy caused Login to loose its docstring and became no longer publishable.

This was problematic because it was not detected by the test suite, so HTML test is changed to use zope publication instead of calling the view on the context.

Another notable point is that this behaves differently in ERP5JS, the Logins were still visible in ERP5JS. Because we don't actually rely on "no docstring on the class" to prevent publication, this is not changed.

See merge request nexedi/erp5!1571

There was opening and closing tags mismatch:
    <div><a href="#references1_anchor">References</div></a>

This seems to be used only in Localizer's ZMI. This is not compatible
with newer version of DocumentTemplate, but because we don't use, it's
better to remove the functionality

Most service worker precache scripts reference a favicon.ico, but this
is using the default favicon.ico from Zope and even though it was included
in all ERP5JS and OfficeJS web sites, this was mostly not used, only
web_renderjs_ui web pages reference favicon.ico.

There's a favicon.ico in erp5_xhtml_style skin folder, but the skin
folder is not in ERP5JS skin selection.

On Zope2, this caused ERP5JS and OfficeJS application use the default
Zope favicon. On Zope4, the service worker can not fill its cache
because of 404 errors.

To solve this, provide a favicon.ico in ERP5JS skin selection, by
copying the one from erp5_xhtml_style. We also reference it explicitly in
web site layout properties so that it remain in the cache.

OfficeJS applications do not use favicon.ico explicitly. They use icons
in their web application manifest, but this does not seem to be use as
favicon unless the PWA is installed. This part is not addressed by this
commit.

Globals.DevelopmentMode initial value is False but it is initialized to True at
Zope startup (App.config.{setConfiguration,DefaultConfiguration}).

zope.conf of Zope2 (zopeschema.xml) used to define event/Z2 log files and
properly configure logging but this is not the case anymore with
Zope4 (wsgischema.xml).

This adds path to event/Z2 log files as command line parameter as we probably
don't need further configuration such as message format (SlapOS recipe has never
handled it anyway).

In order to stabilise UI tests which are waiting for the loading animation
to play during language change.
Pick a small delay so users should not notice (value suggested by Romain).

When we have more than one language and many documents with same reference but with different language, the first impression is that we have many duplicated documents.

Display the language, we can easily detect that we have same reference to different languages.

While in a "single" list field, we need an empty item that will be
selected by default, there's no such need on a multi list field.

These empty items also have the problem that their meaning is not
really clear.

Instead of having empty string items which for which the meaning is not
really clear.

processing_node -3 is not automatically used anymore, such failure would
nowadays be a processing_node of -2. So simplify this comment.