test fixes

[datafeed cw parser] refactor: split logic that was in the parser into: * an "item builder" component, turning an etree xml node into a specific python dictionnary representing an entity * "action" components, selected given an entity, a relation and its role in the relation, and responsible to link the entity to given related items (eg dictionnary) This changes make it easy to add new action or to override existing one's behaviour according to the entity, the relation, etc... Also refactor the xml view so one can also override what's included in the tag used to mark the entity as used by a relation, hopefully allowing to reduce the number of http requests needed for a full datafeed import.

[datafeed, entities] url for entities from a datafeed source should be on their origin site. Closes #1769391

cleanup

[3.10] fix typo and error for pre 3.10 migration w/ postgres: we *must* call init_creating else the fti isn't correctly initialized, which may lead to silently rollbacked transaction

[migration] don't add CWUniqueConstraint if already there (may occurs when running a migration in multiple times

[datafeed] catch all exception, we should continue to process other urls

fix 3.13.1 migration

Added tag cubicweb-debian-version-3.12.7-1 for changeset a07517985136

Added tag cubicweb-version-3.12.7 for changeset 628fe57ce746

3.12.7

[ms, entity metas] add 'actual source' to entities table / base entity metadata cache. Closes #1767090 this is needed since for entities from 'copy based sources' such as datafeed, we want entity.cw_metainformation() to return as 'source' the datafeed source, not the system source (ie the source where the entity is actually stored). For both performance and bootstraping reasons, we should store this information in the `entities` table and in the _type_source cache.

fix 3.13.0 migration script (closes #1766202)

backport stable

fix 3.12.0 migration script (closes: #1766202)

[test] update to 3.13 api

[migration] when there are no data migration needed, we still want to rewrite the configuration and to restart the server. Closes #1760927

[migration] bugfix: c-c list was not displaying the name of a cube requiring another cube without constraining the version.

backport stable

[rqlrewrite] test and fix rql snippets insertion when several snippets match an optional variable

[datafeed, cw.xml] xml now carry entity's source information, interpreted at the other end so that for instance when an entity from elo is seen when importing cwo, it's properly marked as coming from elo source if one exists

[datafeed] actually raise on error

[python 2.5] compat fix

[datafeed, cwxml] parser now try to complete source's urls so one can add url such as <myinstance>/project. Closes #1759908

[server] fix unicode conversion capability in UndoException By overriding the __unicode__ method, we ensure unicode builtin will return Unicode object even for python versions that don't have a __unicode__ method by default (python < 2.6) In previous versions, when this method is not defined, string conversion is attempted, and the result of string conversion is converted to Unicode using the system default encoding which is ascii the most of the time.

[book] add autoload section from code and fix sphinx warnings * drop misleading faq entry since clear explanation can be found in internationalization chapter * use autodata sphinx directive instead of raw text * fill empty chapter with autodoc: - devweb/httpcaching.rst - devweb/resource.rst - devweb/views/idownloadable.rst - devweb/views/wdoc.rst * use shebang in excerpt to denote shell code * fix some bad indented docstrings in javascript files

[controller] fix wrong view name when error occurs in UndoController

[datafeed] renaming parser attribute to parser_id makes things clearer

cleanup and assert we don't import unexpected stuff

[source config] fix synchronization on configuration entity changes