some improvements:
- use xpath expression in test instead of complex eval expression
- use classList instead of className

Also, manually handle mouseover for highlighting relation field menu lines instead
of using css. Indeed, we have to manually handle possibility that highlight could
be done either by mouse or by keyboard.

```py
kw = {...}
kw['metadata[key]'] = 'value'
connector.createSession(**kw)
```

but it would be more beautiful to be able to write it like this:

```py
kw = {...}
kw['metadata] = {'key': 'value'}
connector.createSession(**kw)
```

This would need change in the way the request data is serialised in
createSession.

This fixes things like:
  portal_catalog(source_reference={'query':'foo OR bar', 'key':'KeywordKey'})
or:
  portal_catalog(query=AutoQuery(source_reference='foo OR bar', key='KeywordKey'))

So we get:
  (`...`.`reference` LIKE '%foo%' OR `...`.`reference` LIKE '%bar%')

instead of:
  `...`.`reference` IN ('{\'query\': \'foo\', \'key\': \'KeywordKey\'}',
                        '{\'query\': \'bar\', \'key\': \'KeywordKey\'}')

Also, remove default values as it makes it easy for caller to leave arguments
out by mistake.
Also, use positional arguments as they are faster than named arguments.

See merge request nexedi/erp5!1698

this makes it easier to select the value by copy and paste.
Without the space, when double clicking on "value" in "key:value",
all "key:value" gets selected, with the space, only "value" gets
selected.

Invoking Zuite_waitForActivities when portal_activities is already
subscribed and while another thread from this process is already
executing activities causes infinite "Got a lock error, retrying...".

This situation happens when developer makes a mistake (runs the
tests without unsubscribing activities), so be nice to the developer
so that they don't have to restart Zope.

It's enough to set on the class and if we don't have it on instances, it
saves a few bytes in pickles

There have been a regression recently

Depends on nexedi/slapos!1290

See merge request nexedi/erp5!1695

When installing scripts or components from a business template, set
the _erp5_coverage_filename property so that the plugin is able to find
the corresponding code from the business template filesystem
representation.

coverage might use these to compute the upload URL for coverage data

coverage needs to be started before importing the modules, so it can
not be started so late in runUnitTest, it has to be started before
importing any Products/ERP5* code, ie. in slapos wrapper script.

See https://coverage.readthedocs.io/en/6.5.0/faq.html#q-why-do-the-bodies-of-functions-show-as-executed-but-the-def-lines-do-not
for more details about the rationale

Use super to always call the overridden method.

Extend the API to pass the business template being installed.

Call the hooks also when installing sub objects, they were not called
in one place.

Introduce an action script which redirects to the alarm, so that
"something happen" and user can see that the method was invoked. This
also makes it possible to use this action in selenium tests

With ERP5 workflows, transition id is different (it has a prefix),
we must use transition reference

by using getDocumentValueList to lookup all documents at once instead
of using getDocumentValue for each document

test names are easily longer than the 20 characters that was set
in this field

See merge request nexedi/erp5!1693

This does not seem to be an actual problem because the object is
declared protected, but these missing security declarations make
testSecurity fail.

We have been using `filter(lambda x:'/erp5/' in x[0], error_list)` as a
way to see only missing declarations from ERP5 code, but only ERP5
filesystem code contains /erp5/ in the filenames, in-ZODB components
filename is set to something like <portal_components/document.erp5.FTPConnector>

Change to also check methods for which filename contains <portal_components
and also to print the ignored methods

The logic to guess groupable lines was considering source of
destination side of an accounting transaction depending on
preferences. This was not good, for two reasons:
 - with internal transactions we want to consider both the source
  and the destination
 - if user preferences are mis-configured, grouping would not work.

switch to a logic where we consider both source and destination
sides to prevent these problems.

See merge request !1683

from 45c03413 (comment 171160)

See merge request !1689

This is still needed for now, see 19327cc0

because it contains legacy workflow objects.

because it contains legacy workflow objects.