fix ldapuser test

Prepare 3.15.4

migration: replace empty passwords with random ones on upgrade

ldapuser2ldapfeed: create CWUsers with random passwords, not empty ones

[source/native] only system source users should be authenticated by the native source (closes #2465883)

[web test] fix test broken by 199fefe396ab

[fti] add some documentation

[form] add assertion to ensure field has a name when `role_name` is called

Added tag cubicweb-debian-version-3.15.3-1 for changeset 7ad423a5b6a8

Added tag cubicweb-version-3.15.3 for changeset 0e012eb80990

[pkg] prepare 3.15.3 release

[entity edition] stop comparing values as it may trigger exceptions (e.g. timeseries cube tests) Rely on a changes in yams instead where we have guarantee to have only comparable objects. This follows changesets 1527b012802f, 25da1e5c7c50 and 4ba11607d84a, and should definitly closes #2423719

[tabbed view] make tabs optionaly lazy: lazy=False on the tabbed view trigger synchronous tabs. Closes #2443038 This is useful for tabs holding a self-posted form for instance.

[web form] refactor Form.__init__ so that extra kwargs are set earlier in the process. Closes #2443040 (eg if you want to use them in .session_key)

[test rql completion] dont limit number of completion, else we've impredicable results. Also should tell rql.parse to not print error on stdout This repairs test introduced by dcd9bc1d1bca which closes #2439846

[entity] fix deprecation warnings (closes #2423719)

merge oldstable in stable

Added tag cubicweb-debian-version-3.14.9-1 for changeset 68c762adf2d5

Added tag cubicweb-version-3.14.9 for changeset 9337812cef6b

[ui] provide an autocomplete RQL bar (closes #2439846)

[ajax] improve backward compatibility for redefined ajax functions (closes #2447189) If an ajax function from stdlib (e.g. js_reledit_form) was redefined in a cube, the cube version should still takes the precedence< => check for existing ajax function on deprecated controller _before_ checking ajax function availability on the new one.

prepare version 3.14.9

[composite] apply composite on new enty too (closes #2450679) Before this changesets, automatic deletion did not applied to entity created in the same transaction. This patch remove this filtering in the dedicated operation. The original filtering where introduced by 5d889b4928bb but no rational where found for this changes. The former behavior is seen as inconsistent and creates bug in some cubes.

[storage] relies on event=='deleted' to detect "delete" event (closes #2450680) When we delete a previously edited entity, it *will* have a cw_edited attribute even if the goal is to delete it. The function get an argument that explicitly contains the "delete" event. The code now uses that instead.

[composite] apply composite on new enty too Before this changesets, automatic deletion did not applied to entity created in the same transaction. This patch remove this filtering in the dedicated operation. The original filtering where introduced by 5d889b4928bb but no rational where found for this changes. The former behavior is seen as inconsistent and creates bug in some cubes.

[storage] relies on event=='deleted' to detect "delete" event When we delete a previously edited entity, it *will* have a cw_edited attribute even if the goal is to delete it. The function get an argument that explicitly contains the "delete" event. The code now uses that instead.

[web navigation] previous / next icons are missing mandatory alt attribute

[testlib] consider rset for template / view selection (closes #2447183) The original code did something like: kwargs['rset'] = rset viewfunc = lambda **k: viewsreg.main_template(req, template, **kwargs) kwargs.pop('rset') return self._test_view(viewfunc, view, template, kwargs) meaning that by the time ``viewfunc`` (i.e. the lambda function) is called, ``rset`` was popped from ``kwargs`` and had no chance to be passed to ``main_template``.

[documentation] fix some ReST compilation problems

[test] remove no more necessary attribute cache manipulation to repair test

[entity] more tweaks for entity attribute cache handling on cw_set/cw_create to fix test regression * when an attribute is listed in dont-cache-attrs, purge any previous value * test value is actually modified in EditedEntity.__setitem__ before adding attribute to dont-cache-attrs

[entity/hooks] fix regression introduced by 4ba11607d84a on create_entity, attributes weren't cached, but they are since 4ba11607d84a and this triggered a bug: used values may be transformed by hooks (e.g. TidyHtmlFields), ending in invalid value in the cache. This is the regression. Now, set_attributes used to update the attributes cache while suffering from the very same pb. This is the now fixed old bug. Transformed attributes are now recorded in a transaction shared variable, and those are omitted from the client cache.