backport stable

[test] properly close connections during test

[repo] on repository initialization, properly shutdown intermediary repository

[repo] on repository shutdown, we've to close the new eid creation connection

[auto-reload] we should never call load_module with use_sys=False, that may lead to inconsistency with module interdependancy (eg module get imported by another one, then is reimported by the vreg to load its appobjects. Cleanup of sys.modules done before reloading should be enough.

[rset] do not filter rsets with __getstate__, ensure whatever flies with pyro has no .req attribute, also set the ._rqlst to None since it will be reconstructed later on demand

[security] SchemaRelation.has_perm grows to new arguments allowing to specify subject/object entity type when actual eid are not known, but their types are

[linkto action] refactor to ease overriding to control redirection

[hooks] entity_oldnewvalue should never return _MARKER: if the attribute isn't in the entity dict (hence unmodified), return the old value

Use the qunit automatic launcher for existing testcase.

Add a QUnitTestCase class to run qunit test case.

[twctl] add missing poststop method #1058268

[rset] make rsets picklable (again ?) #1056422

[jstests] nicer test page title

[jstest] get back more tests writen during the sprint by vgodard

cleanup

backport stable

Added tag cubicweb-debian-version-3.8.4-1 for changeset b7883287f40c

Added tag cubicweb-version-3.8.4 for changeset d00d1fab42af

cleanup

[selectors] relation_possible selector accept None as action, in which case no perms checking is done

3.8.4

[xhtml] fix soup2xhtml to deal with malformed div,body and html tags which may leads to malformed return value

[source] remove debug

[entity] add .clear() for bw compat

[3.9] api update (mostly js) + add support for FCKEditor in formContents js function

[json] fix json serialization for recent simplejson implementation, and test encoding of entities as with earlier simplejson implementation, iterencode internal stuff is a generated function, we can't anymore rely on the _iterencode overriding trick, so move on by stoping isinstance(Entity, dict). This is a much heavier change than expected but it was expected to be done at some point, so let's go that way.

[adapters] enhanced warning

[test] fix test broken by 5721:61d6a4caa963, and update it to proper adapters to avoid deprecation warning

[3.9] update warning message

[iprogress] move adapter to entities.adapters

fix wrong calculation of clean_session_interval

[ITree] more fixes, fix fixes

[idownloadable] depends on a new mtconverter release

[ITree] fix adapter/_done_init to handle tree_[attribute|relation], add maxlevel argument to base tree view since one side effect of the new code is that it find tree view for entity types it didn't find before, hence one may want to limit tree view depth when it wasn't necessary before

[3.9] fix tests

[web] move itree adapter to entities, it may be used outside the web interface

[view] add a new entity_call method to entity view protocol, allowing some to work with not yet created entities. Also, start considering 'eid' form parameters where we only consider 'rql', so we can move on bloquing arbitrary rql inputs (more to do on this...)

[idownloadable] refactor IDownloadable primary view to make overriding easier

[web] ajax_replace_url superseed build_ajax_replace_url, more generic and relying on the json controller (closes #750090)

[form] refactor meta-attributes handling: hide them by default using uicfg.afs instead of returning None (or not, according to skip_meta_attr flag argument) in guess_field

[repo] fix typo

[req] fix format_time w/ 0:0 which is evaluated to False

[view] ensure fallback_oid is specified before trying to select fallback view, else simply raise the exception

fix merge, forgot to save a buffer :(

backport stable

[rql2sql] fix special 'X relation NULL' case (used by security insertion), broken by recent removal of IS operator in RQL

[javascript] add a name property to namespaces, mainly for convenience in helpers such as movedToNamespace

[rql] IS operator killed in rql 0.26.2, use equal instead

[security] consider any rql expression refering to X while X is being created as satisfied

[rql] more update to 0.26.2 api

[querier] fix rset description bug with some union queries

[devtools] fix get_available_port implementation: it should only return the port number if unable to connect

[javascript] move CubicWeb object declaration to cubicweb.js, remove CubicWeb.provide/require, use cw instead of CubicWeb

[javascript] move Deferred implementation to cubicweb.ajax.js

[form] fix untested (unread?) has_been_modified implementation...

backport stable

[autoform] fix vreg module name

[pylint] fix detected name errors

[pylint] fix detected name errors

[repo server] fix startup bug, lgc.daemon.daemonize return 1 for the parent process, not -1

backport stable branch

[source] don't create the eid_creation connection at source initiliaztion time This fixes the "already opened connection bug" in db-restore for instance.

[schema view] only catch key errors

[schema view] improve it slightly by attributing the same color to rdef of the same relation type

[rset] Add a line break the first result in repr of multiple rows rset Improve readability of rset while debugging by aligning the first value with the others if the rset have multiple row. This first line was usually not read as it' was on the same line that the first part of the repr ("<rset <RQL>, (%i rows) ...")

[schema diagram] ticket #191599, make schema image look better

backport stable

[workflow image] remove debug code

[request] fix typo

[security] view was giving web access to every fs file accessible by the web server process. Fix this security hole by storing file in session data, not direct form parameters

[hooks] fix typo in workflow hooks category

[security hooks] fix bad merge

backport stable

Added tag cubicweb-debian-version-3.8.3-1 for changeset 1ccaa9247860

Added tag cubicweb-version-3.8.3 for changeset 7e6c6a2a272d

[edit ctrl] fix bad interaction with fields: the loop on actual fields should be done in has_been_modified

[web test] add basic windmill tests and documentation - connection / deconnection - user creation

[web test] Add windmill functional testing capability.

3.8.3

i18nupdate

[editform] consistent [do]reledit between attributes and relations: check if rdef is hidden in main section of uicfg.afs for both attributes and relations, and in the reledit view, not in doreledit

on entity creation, accept attributes without any update access

[table] propagate paginate

no more need for specific handling of table view's displayaction it's handled as other view arguments in vidargs

[rset] do not break when rset.related_entity(x,y) is not an entity

[migration] fix schema synchronization bug sync_schema_props_perms('Repository', syncprops=False) to synchronize permission of entity's attributes wasn't working as expected: only the entity's permission were updated.

[javascript] refactor execution flow of handleFormValidationResponse so that location hash is set to #header whatever the error is

[package] now depends on yams 0.30 because of xy API changes

[xy] switch order of arguments for register_prefix()

HAVING sql clause don't know about comma...

[web/formwidgets] check on list (entity.related_cache swapped from tuple to list a while ago without known reasons), do not unicode wrap the eids

merge

[win32] defer import of logilab.common.daemon The module is not importable on windows, and etwist.service carefully avoids importing it (but imports etwist.server). Wait until we are sure not to be on windows to do the import)

[ajax] refactor/cleanup low-level ajax functions * loadxhtml/replacePageChunck/reload_component/reload_box deprecated in favor of loadXHTML / ajaxFuncArgs * some other cleanups in cubicweb.ajax.js * add js_render which will replace js_component (more generic, nicer argument passing handling)

[querier] test and fix potential bug in insertion query with multiple relations inside, caused by a name conflict. Also fix another bug non reproduced by using a set instead of a list (list may leads to compare entity instances which isn't allowed).

[form] avoid spurious warning subsequent for form's action refactoring w/ autoforms

backport stable

[web test] Add a CubicWebServerTC class to run test with a cw web serveur available.

[web server] vregistry may now be passed as named argument in CubicWebRootResource constructor This will be needed to start an http server for testing purpose. Also, server run command has now a debug name argument to set debug mode explicitly

[etwist] twisted won't install signal handler outside the main thread. This patch avoid some warning about this issue.

[vreg] ensure you can register an object from module not known yet. This ease registration of arbitrary objects during testing.

[web] cleanup use of config in web application initialisation

fix unit tests by not using the new create_eid implementation with sqlite Actually, fix a couple of issues with a previous attempt: * do not use double underscores as method patching will fail with these * old method must always be used with sqlite regardless of the no_sqlite_wrap attribute

cleanup the no_sqlite_wrap attribute after this test is executed

merge back on-site fixes

various brown paper bag fixes

merge

fix create_eid for sqlite (and bring back tests)

backport improved on-site change

TimedCache now only accepts values expressed in seconds updated ldapuser.py and pyrorql.py to that new interface.

backported to stable some changes made on site for a customer

logging settings * document helpful log format when dealing with MT issues * on Win32, use a logrotate equivalent so that logs get a bit lighter (requires updates logilab.common)

[win32] fix deadlock occuring on the sequence tables with SQLServer actually, this deadlock would occur with any db backend other that PostgreSQL as the previous code was heavily relying on PG's SEQUENCE facility, not available elsewhere. Deadlock description: Thread1 starts creating entities (and therefore calls create_eid): -> this creates a DB-level lock on the entities_id_seq table, which will last until end of transaction Thread2 calls create_eid, which acquires the Python lock object, but updating the entities_id_seq is held by the DB lock Thread1 wants to create a new entity, calls create_eid, and is stuck by the Python lock object held by Thread2. Solution: use a separate connection to read and write the entities_id_seq table.

log the user name on session opening for easier debugging

more robust handling of looping task configuration in ldap source * use minimum times * updated documentation and help strings * fix the minutes vs. seconds confusion

backport stable

[rql2sql] fix scope computation of variable from temporary table

stop being dumb please

backport stable