/reviewed-on !739

/reviewed-on nexedi/erp5!740

zope.conf default encoding is also used for responses and converters, so
for consistency we should also force default encoding there.
This is especially useful when running tests, because zope.conf is not
used and default values are used.

https://tools.ietf.org/html/rfc4329#section-7 states that text/javascript is obsolete.

Doesn't really matter in the end because WebPage_viewAsWeb forces the content type.

In HTTP/1.1 , when not specified , charset detault to ISO-8859-1

https://tools.ietf.org/html/rfc2616#section-3.7.1

> media subtypes of the "text"  type are defined to have a default charset value of "ISO-8859-1" when   received via HTTP

Zope adds a charset with default encoding (
https://github.com/zopefoundation/Zope/blob/2.13/src/ZPublisher/HTTPResponse.py#L512-L515 ) but does this only for text/ type, but we serve our javascript as application/javascript

Correctly handle multiple interface declarations

Reduce DOM modification.

Load gadget in an iframe to improve isolation

Delay gadget creation to prevent creating hundred of iframes in parallel, which slow down the browser

Prevent submitting the form if the relation field contains an unknown search string.

_reindexOnCreation API does not expect an expanded activate_kw, but
really only an expanded reindex_kw - which may contain its own
activate_kw, over which our activate_kw is supposed to take precedence
(see Products.ERP5Type.Base.Base._getReindexAndActivateParameterDict).
It's all so easy !

activate_kw was not honored in immediate indexation code path.

Also, tix typo in comment.

Revert "Folder: call activate on the object instead of portal_activities.activateObject in _recurseCallMethod()."

This reverts commits 3d3ec3cb ((+ fixup), which
broke recursion to objects that don't inherit ActiveObject, as explained in the
docstring.

I also think it was wrong in that default activate parameters disappear at the
end of the transaction whereas _recurseCallMethod may activate itself to avoid
creating to many activities at once. IOW, it's important that created
activities are the same no matter where _recurseCallMethod splits the browsing.

This change needed more review, as well as a unit test.

Folder: call activate on the object instead of portal_activities.activateObject in _recurseCallMethod().

so that activate_kw set on the object is respected.

The site shall be used with default BT5s.

Previous field names came from obsoleted erp5_core_proxy_field_legacy

This was the case before
  commit b41dc8e0
  Author: Vincent Pelletier <vincent@nexedi.com>
  Date:   Wed Sep 12 04:00:08 2018 +0200

      SQLCatalog_deferFullTextIndex{,Activity}: Use serialization_tag .
but I misread the code: when this exception triggered, document was neither
added in parameter_dict nor in failed_path_list. It should be now clearer.

This reverts commit fedebf37.

Finally, it was an issue in selenium runner software, the wrapper
was missing

Also, why make only 2 queries when I can make 2n+1 queries for the same
result ?

getObject is generally a shortcut for "return self", and on brains for
traversing to corresponding document.
Except... Not everywhere. portal_catalog for example defines a completely
unrelated method which takes a UID as argument and does a query retrieving
one document. Which is basically totally redundant with its __call__
method. Still, of course, it is used in several places.
Also, ERP5Form.Report defines yet another getObject method, taking an
acquisition context as argument and traversing to some other path.

So do not use getObject (at least until the code recovers some sanity and
consistency), but getPath + custom activity API. Oh well.

Just like regular indexations, fulltext indexations are subject to last-commit-wins.
Which means that it is possible to reach a state where the fulltext table is
persistently desynchronised from ZODB:
- start fulltext indexation activity on many documents (typically: 100)
- modify one of the documents being indexed
- start fulltext indexation activity caused by this edit, and assume indexation only
happens for this object
- commit the single-object indexation (because it is very fast to retrieve fulltext
data from just one document)
- commit the many-objects indexation later (because it is much slower to
retrieve 100 fulltext representations)

As a consequence, cod emust spawn one fulltext indexation activity per
document, each with the appropriate serialisation tag. Serialisation tag which
must not conflict with regular indexation, so use a fixed prefix.
As a consequence of having to spawn one activity per document, use a
grouping method to still index by batches to amortise transaction overhead.
Keep the same method_id as before for backward-compatibility (maybe
dependencies on this value exist, even though it is bad practice).
Rewrite SQLCatalog_deferFullTextIndexActivity so ot works as a grouping
method, simplifying it in the process:
- build parameter_dict with all entries, as we already know all needed keys
- None is not callable, so test "not None" in just one expression
- remove whitespace at end of line
- use GroupedMessage API