i18n update (closes #4826490)

merge 3.19.7 into 3.20

merge two default heads

Added tag cubicweb-version-3.19.7, cubicweb-debian-version-3.19.7-1, cubicweb-centos-version-3.19.7-1 for changeset ac4f5f615597

[pkg] 3.19.7

Added tag cubicweb-version-3.20.0, cubicweb-debian-version-3.20.0-1, cubicweb-centos-version-3.20.0-1 for changeset 7e6b7739afe6

[pkg] prepare 3.20 release

[skeleton] simplify debian rules file

[migration] don't crash when deleting a non-existing rtype When asked to delete a relation type, instead of checking if it was computed or not, just delete both the CWRType and CWComputedRType.

[debian] include "metapackage" in the cubicweb package's description Lets lintian know it's ok for it to be empty.

[debian] ensure all packages use ${misc:Depends} and ${python:Depends}

[debian] update cubicweb-dev's lintian override Switching from python-support to dh-python changed the installation path.

[server] add a db-namespace option in source definition (closes #1631339) "namespace" is preferred to the postgresql term "schema" to try and avoid confusion with cubicweb's schema. cubicweb now depends on logilab-database >= 1.13.0

[serverctl] add db_transaction context manager

[server] refactor cleanup() code used by "cubicweb-ctl delete" Try to make each logical step (i.e. "delete database" and "delete user") a bit more visible and explicit. This is a preparatory work to easily plug the "delete namespace" feature.

[views] visual refactorings (closes #3803685) - pageContent: - remove the shadow - add a light border - slightly round the corners - left column: - ensure width make it aligned with the logo (header) - ensure the top border of the first box is aligned with the pageContent div - do not use serif font for box headers

[web] everything old is new again Kill the "new" cubicweb.css, and move cubicweb.old.css back in its place. Closes #4763360.

[pkg] bump version to 3.20.0

[views] use icons for user menu and login (header) (closes #3803684) - use icons from the embedded pictograms font for: - authenticated user menu (userActionsBox) - login (CookieLoginComponent) - do not display "anonymous" when not logged in

[utils] provide a function to return an admin connection from an appid Closes #4723580.

[debian] Install wsgi directory in cubicweb-web package Closes #4752375.

[test] fixes to make tests pass with recent lxml lxml (on jessie at least, ie. 3.4.0) does not remove endlines anymore.

[server] Remove useless 'is None' checks 'pb' is never none since 3386fd89c914 when support for the 'APYCOT_ROOT' env var was removed.

[security] check attributes: dispatch on the "add" action if entity was just created cw_set on a just-created entity (i.e. created in the same transaction) should behave the same as setting the attribute directly on creation: check the 'add' permissions, not 'update'. Closes #4740310.

[doc] Add BigInt to the list of built-in types

[devtools] call turn_repo_off in tearDown (closes #4673491) lets it catch leaked sessions even when running a single test

[book] typos / clarifications in create-instance section Clarify that cubicweb-ctl upgrade needs to be run whenever cubicweb or a cube is upgraded. Related to #2632425.

[repository] don't mangle the stack trace on exception

[pkg] Make twisted recommended only It's not necessary when using wsgi or pyramid. Closes #4639508

[devtools] Use the pause_trace context manager instead of pause_tracing/resume_tracing The later were deprecated in logilab-common 0.63.1

[serverctl] rename remove_cube to drop_cube (closes #4545093) For the sake of consistency, since commands are currently named: - add_{cube,entity,attribute,relation} - drop_{entity,attribute,relation} - remove_cube

[wsgi] add tornado http server

[wsgi] add message about the active interface Similarly to what done in "stdlib".

[cwctl] rewrite wsgi method choices This will help devs to add new choices by adding their callback to WSGI_CHOICES. Note: ``options`` becomes a property because ``wsgichoices()`` should be evaluated at the last-minute.

[skeleton/debian] use dh_python2 instead of dh_pysupport for new cubes The former is deprecated.

[web] Add source for jquery.qtip.js Downloaded from https://github.com/Craga89/qTip1/raw/master/1.0.0-rc3/jquery.qtip-1.0.0-rc3.js Related to #3851121.

[web] Replace minified copy of jquery.flot.js 0.6 with non-minified version Related to #3851121

[web] Update jquery-treeview to the latest version Files copied from commit 61f230828ace1b54c2cc54bc549ece261b11842e of https://github.com/jzaefferer/jquery-treeview

[web/css] move jquery.treeview.css override to a separate file Patched embedded code copies are a maintenance disaster (even more so than plain embedded code copies). Let's not do that.

[web] Stop patching jquery.treeview.js This logically reverts part of e9b7cd2e9012 "allow treeview to work correctly in a tab #345293". The treeview and ajax code have suffered a number of changes since then, this change doesn't seem to be necessary anymore. It is most likely unneeded since f65208c9dbdc "[javascript] use jQuery.one('ajax-loaded') instead of jQuery.bind() in add_onload to avoid multiple callback executions".

[web] Beautify jquery.treeview.js Fully mechanical, just whitespace changes.

[entities/wfobjs] add missing `DBG_SEC` debugging informations and a new `transition` capability

[warnings] put an end to warnings in the sqlite driver over `str` being sent instead of unicode strings

[web/request] clearly mark user_callback-related methods as deprecated The deprecated decorator not only emits a warning at run-time but above all makes the deprecation immediately apparent in the source. closes #3567793.

[book] document __unique_together__, remove bad RQLUniqueConstraint example RQLUniqueConstraint should be avoided whenever possible. Related to #3753250.

[schema] stop using RQLUniqueConstraint (closes #3753250) The last uses are replaced with unique together constraints.

[dataimport] Have ucsvreader's API match that of csv.reader (closes #3705701) 'separator' and 'quote' are replaced with 'delimiter' and 'quotechar'.

[utils] Add a '_cwtracehtml' GET parameter to trace self._cw.w() calls (closes #4601327) The core of this patch is in UStringIO.write(). When tracing is enabled, write() doesn't just append the 'value' argument to the underlying list. Instead, a stack trace is recorded and a special HTML "source" is formatted. The output with tracing enabled is an HTML page, with the original HTML escaped, and made clickable to show the stack trace when the write() call was done. This allows answering the recurring question: "who wrote this tag here?!"

[datafeed] Add a raise_on_error parameter to DataFeedSource.extid2entity And pass the option from various `process_` methods in existing parsers. This makes debugging in tests easier. Closes #4601191.

[test/view forms] use the official, undeprecated API

[rset] kill the rset._rqlst cache Right now it "works" for the standard, internal uses. However when we will fold ClientConnection and Connection, it will hurt, because suddenly we get more cache hits, and the following situation would become commonplace: * there is an un-annotated _rqlst given by the querier * some view (e.g. facets) requests the .syntax_tree, which takes a copy of _rqlst * the view actually expects the rql syntax tree to be annotated, but it was not, hence we crash. Related to #3837233.

[source/native] session -> cnx Also swap process_results arguments order to make cnx come earlier.

[hookstests] change got/expected order

[config] kill a getattr

[tests/web] switch previous commit (136b5f995f8) to the new test api Related to #3670209.

[web] remove unused argument from treeview._init_headers

[doc/3.20] mention CWEP-002

[web] stop accepting the magic __message form parameter This has been deprecated for a while, and replaced by _cwmsgid, which doesn't allow arbitrary content.

[web/request] add security_enabled method Just forward to the underlying repo connection. This can be useful in the ORM when dealing with a "code-only" attribute, i.e. something that shouldn't be directly user-visible but must be accessible from a web request.

[dataimport] Fix use of _create_copyfrom_buffer() (related to #3845572) This is used indirectly by SQLGenObjectStore when running on a PostgreSQL database. Regression introduced by commit 70056633085c.

[c-c configure] make it possible to specify the section for sources configuration (closes #3477678)

[server] fix 'cnx' variable confusion in DBG_SQL exception case The rollback handling expects 'cnx' to be the cubicweb Connection, but the DBG_SQL block was replacing it with an sql connection, leading to lulz down the line. Also remove obsolete getattr (the sqlite wrapping is now done at the cnxset level, so cnx.cnxset.cnx should be the right thing already).

Provide sufficient context to check 'delete' permission in AjaxEditRelationCtxComponent Call rdef.check only when both fromeid and toeid are available. Though only call it once (for the first encountered related entity). Factorize a bit to keep handling of CSS/JS addition the same. Closes #3670209.

merge 3.19.6 into 3.20 branch

Added tag cubicweb-version-3.19.6, cubicweb-debian-version-3.19.6-1, cubicweb-centos-version-3.19.6-1 for changeset 934341b848a6

[test] missing unittest.main() call in unittest_http_headers.py

[test] make unittest_webconfig independant of CWD if apphome is unset in ApptestConfiguration constructor, this later looks for a 'data' directory in CWD, making this test runnable only from web/test (and pytest do chdir before executing tests, so we did not noticed this problem).

[webctl] do not ask questions if verbosity is at 0 Closes #4641960.

[web] Slight tweak on jQueryUI's colors on links (related to #4564046) Follow up to commit 16f550b48d57 which was somewhat brutal. This change only overrides jQueryUI's CSS rules inside tab panes and leaves actual tab buttons alone.

[view] Remove unused imports

Remove obsolete __future__ imports Generators have been available since 2.3, "with" statements since 2.6.

[devtools] Remove unused local function Unused since cw's first commit.

[uihelper] Fix a possible NameError in meta_formconfig could occur when using FormConfig with a custom uicfg object. Closes #3802298.

[devtools] Avoid db_cache collisions and mis-loading This is done by stopping to share the db_cache between TestDataBaseHandler instances It may seem overkill, but is the only way I could get it to work properly. I think the whole db caching code needs some rework to better work with different db handlers. Closes #4601328

[devtools][pg] Remove a wrong assertion in _backup_database When using pre_setup_database, a repo gets initialised before calling backup_database. Hence self._repo is NOT None it that case.

[pkg] prepare 3.19.6

[merge] backport 3.18 fixes

Accept '==' operator in cubes dependencies '==' is the operator used in the python dependencies specifications. See PEP 440 http://legacy.python.org/dev/peps/pep-0440/#version-specifiers Using the same operator in cubes dependencies allow to make setuptools handle the cubes dependencies while keeping cubicweb happy about the dependencies. Closes #4375144

[session] don't silently ignore commits commits while in the precommit or postcommit phase are silently ignored. This is OK in precommit since there'll be a commit afterwards anyway; OTOH in postcommit it may surprise the developer who didn't read the documentation for postcommit operations carefully enough. Related to #4383922.

[migration/bootstrap] add explicit index removal for sqlserver, be tolerant to migration replay Closes #4618944.

Added tag cubicweb-version-3.18.7, cubicweb-debian-version-3.18.7-1, cubicweb-centos-version-3.18.7-1 for changeset cb96f4403cf2

[pkg] prepare 3.18.7

[migration/3.18] speed up the migration and be a little bit more informative Closes #4619277.

[schemaserial] be robust against duplicated CWUniqueTogetherConstrain entities arising from bug in the pre 3.18 era Closes #4619278.

[server] Handle unique constraint violations under recent sqlite The error message changed from "columns foo, bar, baz are not unique" to "UNIQUE constraint failed: table.foo, table.bar, table.baz". Closes #3564510

[merge] backport 3.17 fixes

Added tag cubicweb-version-3.17.18, cubicweb-debian-version-3.17.18-1, cubicweb-centos-version-3.17.18-1 for changeset cda4b066933f

[pkg] prepare 3.17.18

[source/native] handle newer Integrity error messages from sqlite > 3.7 Similar to a0cf2993b6d3 that was done for 3.18.

[source/native/backup_restore] have a unique tunable blocksize for the dump phase Thus we make memory errors much less likely. Closes #4618949.

[urlpublishing] handle sub-types in RestPathEvaluator Since commit 31a1813d53f3 "[entity/url publishing] fetch_rqlst should use 'is_instance_of' instead of 'is'", the RestPathEvaluator's input rset can have more than one etype, so it can't always use sameetypelist. Related to #3825488

[views] fix ecsvexport selector introduced by e48e5a597ccc Fix the view and make web/test/test_views.py pass again.

Add custom checker for Password values We override the default converter to pass Binary values through, but don't do anything about the checker. This worked previously because yams allowed StringIO instances, although its converter didn't do the right thing for them. Fixing this in yams requires that we properly register a checker.

[component] give Links a __repr__

[debian] move markdown dependency from cubicweb-web to cubicweb-common Pointed out by David Douard.

[web/auth] stop playing games with locals() We only want to know if any retriever found something.

[entity] rather than crashing with an IndexError on entity creation, raise a meaningful exception

[web] partial backout of #8391bf718485 to restore RelatedObjectsVComponent Apparently the intended deprecation did not work, and this component is used in at least the tracker cube.

[server] typo fix

[devtools][pg] Activate the non-durability settings of Postgresql. It speed up the PG tests by a factor of 3. See http://www.postgresql.org/docs/current/static/non-durability.html.

[server] fix LoginPasswordAuthentifier.authenticate some kind of typo there.

[web] Override jQueryUI's colors on links (closes #4564046)

[hooks] Fix precommit event logging message

[service] allow repo_stats for users It's used by the siteinfo view, which is available to managers and users. This change prevents a crash in that view.

[views] Escape class attribute value in CWGroup incontext view

[devtools] use a specific test_db_id when disabling anon The anonymous user is created (or not) in postcreate, which for tests means when creating the cached empty database. The anonymous_allowed=False test classes should thus not share their template db with other test classes, otherwise either they end up with an unexpected anonymous user, or the others miss theirs and things fall apart.

Restrict yams version CW 3.19 is not compatible with yams >= 0.40

[facets] Honor 'start_unfolded' facet attribute (closes #4502799)

[facets] Correctly look for inputs of type "hidden" (closes #4502768) jQuery ':hidden' selector looks at CSS visual properties (eg, 'display', 'visibility'). The intent here was probably to look for inputs of type "hidden", which many facets use to store user selection data (eg, FacetRangeWidget). The problem is that regular text inputs (eg the "has_text" facet which has a '<input type="text"/>') will be picked up by this selector if they are inside a folded facet. Chaos and destruction ensue.

[web] Cache results from 'i18n' ajax controller (closes #4487856) This is a simple transient cache which will be forgotten as soon as the user leaves the current page.

[web] There is no global noop(), use jQuery's instead (closes #4487832)

[server] Add missing import of logilab.database Closes #4469407.

[migration] hackish black magic to bootstrap addition of formula attr during migration Turns out we can't add an attribute to CWAttribute before the attribute exists. add_attribute generates rql with a 'formula' relation, which CW doesn't know about yet, so things get unhappy. Once that is fixed, we need to make the CWAttribute addition operation deal with CWAttribute entities that don't have a formula yet. Finally, move the migration to bootstrapmigration_repository so it happens early on, and use add_entity_type rather than add_relation_type for CWComputedRType.

[server] commit after serializing schema Use separate transactions for schema serialization and postcreate scripts execution, so that the latter can rely on the schema being persistently stored. This matters e.g. if something in a postcreate violates a unique_together constraint, we need the constraint persisted in order to raise a meaningful error. See #4525069.

[book] fix docstrings to please sphinx example rst roles in docstring must be protected (in a literal).

[book] several fixes in the rst files - prevent duplicate markup definitions, - use existing Workflow markup instead of redefining one (in baseschema.rst) - include dataimport - fix non-existant directive autodocfunction - remove the empty 'embedding.rst' file

[book] activate the viewcode extension Automagically make link to class/modules/etc. it finds in the documents.

[book] new theme based on pyramid theme (closes #4291287)

Remove unused lgc.interface imports and leftovers

[repository] provide a .new_session entry point The current .connect only returns a `sessionid` which must be given to repo._get_session(...) to get the real session object. Related to #4151635.

[devtools/testlib] Use actual 'admin' user configuration values Previously ``db-user`` and ``db-password`` configuration values were used to fill ``admlogin`` and `admpassword`` attributes of CubicWebTC. The correct data for these should read from `[admin]` section. Closes #4065070.

[forms] closes #2437859 - Detect and prevent concurrent edition of the same entity. Add the timestamp of form generation to each entity's meta-information fields. On form validation, check that no concurrent change is overwritten and raises a ValidationError in case of concurrent change. A nicer handling with a message and a link to the new version of the entity would be a good thing...

[migration] make sure the repo knows about all constraint types Adding new constraint types is cumbersome. It's easy to forget to do it in a migration script. So sync the constraint types at the beginning of each migration. Closes #3724892

[migration] stop caching the mapping from constraint type name to eid It's not so frequent that a cache seems necessary, and we were not invalidating that cache when adding a new constraint type. Related to #3724892.

[web/json] an empty rset is just fine for ejsonexport The ejsonexport view can just return an empty list for an empty rset. Closes #4005518

[RichString] Add markdown support Supporting markdown requires the 'Markdown' python packages, and since it widely available on the target platforms adding it as a strong requirements will not be a big constraint. Closes #3814302

[ldapfeed] Reduce default value for user-attrs-map option (closes #3824889) This is needed because lgc.configuration.Configuration does not allow removing key/value pairs from the default (due to its use of dict.update() internally). Since CWUser.login is required, users of the add-source command will always be able to override it.

[serverctl] Ask for URL when adding a new source (closes #3824868)

[serverctl] Ask for parser type when adding a new source (closes #3484231) Remove 'quick_start' option to load all AppObjects, including feed parsers.

[doc/3.20] more details on removed code related to #3799117

remove 3.11 bw compat Closes #3799117.

remove most 3.10 bw compat Related to #3799117. The boxes and entityvcomponent objects cannot really be removed as they are still used throughout the code base (and possible cubes). This may be caused by a non-working deprecation and needs some more work.

remove leftover pre 3.9 deprecation warnings

[utils] Remove function-in-the-middle call

[web] set Vary response header to "Accept-Language" when using content negotiation This is slightly annoying because the response actually only varies based on the language we decide to send, which has much fewer possible values than Accept-Language, but that's not in the request, so we can't easily use it. Deployments using varnish or similar and controlling the set of available languages will likely want to override this to allow reasonable amounts of caching. Closes #2105812

[web/cors] don't overwrite other Vary headers Vary is a list of request headers, we shouldn't override others.

[web] add support for HttpOnly cookie flag And use it for session cookies. Closes #4142521.

merge 3.19.5 into 3.20 branch

Added tag cubicweb-version-3.19.5, cubicweb-debian-version-3.19.5-1, cubicweb-centos-version-3.19.5-1 for changeset 3ac86df519af

i18n update 4 strings have disappeared.

[pkg] 3.19.5

[server] hold connection to the db in tx_actions We can be called without a cnxset (e.g. from repoapi).

[wsgi] If multipart cannot parse the POST content, let it pass. multipart can only parse html form data. It the content_type is, for example, "application/json", get_posted_data should not fail but just stop trying to read the content. Closes #4421845

[devtools] Fix JS tests' HTML code

[devtools] "Keep" some temporary files/dirs around to help with debugging The whole QUnitTestCase runs with an @with_tempdir so it's redundant anyway.

[devtools] Fix Firefox launcher in QUnitTestCase (closes #4294727) The main changes are: - stop creating the profile, firefox will create it - point firefox to a profile directory instead of giving it a profile name (this has the added bonus of not polluting the user's profile list) - start firefox once and kill it 5 seconds later to let it finish its profile creation (along with system-wide extensions setup)

[devtools] allow cross-origin requests for qunit We have a mix of file:// html and http:// ajax calls. Which we should at some point fix to all be http, but. Related to #4294727.

merge 3.19.4 in 3.20 branch

[cors] Fix CORS headers generators The Access-Control-Allow-* and Access-Control-Expose-Headers are response headers, not request headers. Hence, we need generators for them. Closes #4412575.

[wsgi] Fix posted files filename reading The filenames are parsed by multipart.parse_form_data, which does the unicode decoding. Trying to re-decode the filename was leading to an error.

[pkg] Depend on Pillow instead of PIL The Pillow library is becoming the de-facto replacement for PIL. It also is way simpler to install with pip than PIL. Closes #4411354.

Added tag cubicweb-version-3.19.4, cubicweb-debian-version-3.19.4-1, cubicweb-centos-version-3.19.4-1 for changeset c4e740e50fc7

[pkg] 3.19.4

merge 3.18.6 into 3.19

Added tag cubicweb-version-3.18.6, cubicweb-debian-version-3.18.6-1, cubicweb-centos-version-3.18.6-1 for changeset d91501356742

[pkg] 3.18.6

[hooks/security] allow edition of attributes with permissive permissions If an attribute has more permissive security rules than the entity type itself, we should be green and not deny action because of an early global entity permission check (with the more restrictive rules). Only if one attribute with the entity-level permission rules is edited will the global check be performed. Note: * the "if action == 'delete'" check at the entry of check_entity_attributes is a guard for a condition currently not happening in cubicweb itself (but application hooks could conceivably call this function with a 'delete' action) Closes #3489895.

Almost backout afcd46716d6a which breaks _select_best raising an ambiguity exception in debug mode. The problem is, before afcd4, *tests* ran in debug mode and we want this (e.g. we want to show, rather than swallow, select ambigüities). We juste replace the bogus __init__(vreg.config) by __init__(True), which is practically equivalent and also much more clear.

[server] fix anonymous_user predicate in tests devtools' TestServerConfiguration overrides the anonymous_user method, but not the anonymous-user config option, so testing for the latter would give the wrong result. Closes #3996664.

[entities] cw_rest_attr_info() should only consider required attributes (closes #3766717) This prevents CW from choosing unique but non-required attributes. None/NULL is a poor choice for RESTful URIs.

[views] csvexport accept an empty rset (closes #4236928) When you tried to apply the 'csvexport' view on an empty rset, the view couldn't be selected and you got a HTTP 500 error. Also add two new test cases.

[views] Display attributes in entity creation form based on "add" permission Previously, the "update" permission was used. Hence in case the latter is more restrictive that the "add" permission, an user may not be able to set such an attribute, despite she may have "add" permission on it. As a result of the change of permissions action in `editable_attributes` method (add/update depending on whether the entity is being created or modified), the "eid" attribute would have shown up in the edition form. To avoid this, it is moved in the "hidden" section (where it should arguably belong anyways). Closes #4342844.

[datafeed] Commit after all deletions in datafeed parser This avoids misleading validation error because schema constraints could be temporarily broken depending on the deletion order. Closes #4372127.

[schema] CWComputedRType is a schema type Hide it from the default views.

[doc] proofreading CWEP002 section

[CWEP002] document computed relations and attributes Related to #3546717.

[CWEP002 migration] support sync_schema_props_perms for computed attribute

[CWEP002 migration] support drop_relation_type/drop_relation_definition/drop_attribute for computed attributes

[CWEP002 migration] support add_relation_type/add_attribute for computed attributes Related to #3546717.

[CWEP002] properly handle serialization of computed attributes Until this cset, only serialization of computed attribute's formula was handled (thanks to the relation definition's properties dict). We now: * properly serialize/deserialize attribute's formula * test first if the database has been migrated and contains related column or not Related to #3546717

[CWEP002] Add support for computed attribute synchronization Related to #3546717. Test and handle the behaviour with several formulas and identified use cases. To do so add a birth year and a computed attribute age to Person in the test schema.

[CWEP002] Add schema finalization checks for computed attributes ``finalize_computed_attributes`` essentially checks that computed attributes types match with the type declaration of the attribute. Related to #3546717.

[CWEP002] Account for attribute formula in schema bootstrap This isn't enough to have computed attribute support, it is only in order not to crash when yams 0.40 is used. Related to #3546717.

[CWEP002 migration] support sync_schema_props_perms for computed relations

[CWEP002 migration] support add_relation_type for computed relations Related to #3546717.

[CWEP002 migration] support drop_relation_type for computed relations Related to #3546717.

[CWEP002 migration] properly raise exception on (add|drop)_relation_definition for computed relation Related to #3546717

[hooks] do not abuse of inheritance for CWRType hooks Just copy the selector to ease readability.

[schema serial] add some comments

[schema] define full_rql on RQLExpression class without it, we may have an error while displaying error which occured during some expression initialization because this attribute is not yet defined while used in the __str__ method.

[schema] properly raise BadSchemaDefinition when some expression mains variables may not be guessed

[CWEP002] properly handle serialization of computed relations We now: * have CWComputedRelation in the bootstrap schema to store computed relations * properly serialize/deserialize it * test first if the database has been migrated and contains the related table Related to #3546717 [jcr: adjust unittest_querier to pass with the added entity type]

[CWEP002] Plug the computed relation rewriter in the querier Related to #3546717.

[CWEP002] refactor rql read security checking Split 'check_read_perms' into 'check_relations_perms' which checks relations 'read' permissions and 'get_local_checks' which build dictionary of local security checks (rql expression) for variables. This allows to check relations 'read' permissions earlier in the process and so to prepare insertion of the rql rewriter: we want to check permissions of the computed relation, not permissions of relations introduced by the associated rule, to conform to the CWEP. Related to #3546717

[CWEP002] introduce RQLRelationRewriter Refactor existing RQLRewriter for later reuse of rewriting relation as specified by CWEP002. Work is different because we simply want to replace a relation by another rql snippet and we don't have to bother with EXISTS, subqueries and all. This rewriter is not yet plugged into the querier. Depends on yams 0.40 API. Related to #3546717

[CWEP002] Add schema finalization checks for computed relations (rules) Related to #3546717

[schema] add utility function to build a CubicWebSchema from a namespace Will be useful for testing computed attributes and relations.

Fix test migration crash waiting to happen due to inferred relations Avoid crashing when a relation definition which is already known (due to being inferred) gets explicitly added. As explained in a comment, this would deserve more thinking, but at least this fixes a test failure with computed relations.

fix typo in syncschema hooks

[test] Make test_undo_api less random The order in which hooks are run is not predictable if they have the same 'order' attribute, which is the case for SetOwnershipHook and SetInitialStateHook. So don't assume in_state will be set before created_by.

merge 3.17.17 into 3.18 branch

[wsgi] Fix unicode decoding in POST The application/x-www-form-urlencoded sent by firefox or chrome is utf-8 encoded BEFORE being urlencoded. Hence, decoding must urldecode THEN decode the utf-8 string, and not the other way around.