During some tests, the runner performs sequences of DUPLICATE and DROP
on databases. Each DUPLICATE is followed by some operations then a
DROP.
Problem is, after a DROP the next DUPLICATE RPC query will attempt a
connection to the just-dropped database, openerp.sql_db will log an
OperationalError ("Connection to the database failed" since the db it
tries to connect to was removed) then the HTTP dispatcher will just
switch to "nodb" and continue on its merry way.
Mute the sql_db logger while attempting to fetch the ir.http object.
Also assert that the logger name is a string in mute_logger, as
`mute_logger(openerp.sql_db)` is syntactically valid, but will not do
anything.
bzr revid: xmo@openerp.com-20140123115413-ax5patcomdfp3ue3
Issue introduced in
revid:chm@openerp.com-20130828161130-641xsmbr8xm53xjx: context of a
fields_view_get can provide a *_view_ref name (e.g. tree_view_ref),
whose value is an xid for the view to use (can be used to specify a
view instead of providing a view_id, mostly for embedded views where
there is no way to provide a view_id).
The view_ref being an external id, it must be dotted. Historically,
undotted xids were silently ignored and the system just skipped to
getting the default view for the model (lowest priority view where
inherit_id is null).
In revid:odo@openerp.com-20130821122955-8c9z0mi8cu48rne3 this behavior
was altered to emit a warning when ignoring a malformed view_ref. When
this change was merged into trunk-website-al, the conditional
(view_ref and '.' in view_ref) was split without consideration for the
semantics change: a syntactically invalid view_ref would fall off the
first branch (view_ref) instead of taking the second one (not
view_ref), the code would not fetch the default view for the model and
would instead generate one (through _get_default_%(viewtype)s_view).
However, closer reading reveals the code was already broken in a
specific case of a syntactically valid but unknown xid (view_id
wouldn't get assigned in the first branch, but the code would not take
the second one, resulting yet again in a view generation even if the
database contains a valid view for the model and type).
Fixed by reintroducing explicit second check on view_id, instead of
merely conditional alternative.
bzr revid: xmo@openerp.com-20140123075742-603n2zthqhxee07y
[CLEAN] res_config, res_users: cleaned call to imd.get_object() + code cleaning
- res_config: raise by default, no need to set the argument at True
- res_users: directly embedded get_user_groups_view code inside update_user_groups_view.
The try / except is still necessary, because when installing a new db, groups and users
are created before the user_groups_view view effectively exist. Due to some circular
references, the try / except is therefore necessary to install a new db.
[CLEAN] ir_qweb: simplified and cleaned QWebcontext implementation
- now using openerp.tools.safe_eval, instead of a custom eval with custom builtins
- removed undefined_handler, hardcoded to a lambda function that returns None for
a missing attribute
- cleaned code, removed now dead BUILTINS, removed commented code
- tools.safe_eval: added a parameter locals_builtins. This allows to copy the globals builtins in the locals. This modification is due to the fact that the locals always returns None, allowing to simplify templates. Otherwise we would have to test the existence of each variable before actually using it. However as the locals always return None for every key, the globals are never checked. Copying the builtins inside the local allows to have a complete locals, but slightly break the globals/locals separation.
[CLEAN] Misc
- setup.py: reverted website-al unnecessary change
- oe: reverted website-al unnecessary change
[DOC] renamed misc_qweb.rst file into ir_qweb.rst
bzr revid: tde@openerp.com-20140116184144-fd5h4a94we13mm72
- now using openerp.tools.safe_eval, instead of a custom eval with custom builtins
- removed undefined_handler, hardcoded to a lambda function that returns None for
a missing attribute
tools.safe_eval: added a parameter locals_builtins. This allows to copy the builtins
in the locals. This hack is due to the fact that the locals always returns None, allowing
to simplify templates. Otherwise we would have to test the existence of each variable
before actually using it.
However as the locals always return None for every key, the globals are never checked.
Copying the builtins inside the local allows to have a complete locals, but sligtly
break the globals/locals separation.
To be reviewed and approved.
bzr revid: tde@openerp.com-20140116182750-1rnw8iljt5a9gb4u
- res_config: raise by default, no need to set the argument at True
- res_users: directly embedded get_user_groups_view code inside update_user_groups_view.
The try / except is still necessary, because when installing a new db, groups and users
are created before the user_groups_view view effectively exist. Due to some circular
references, the try / except is therefore necessary to install a new db.
bzr revid: tde@openerp.com-20140116134056-nyiybx7zsr8rhkiv
For fresh databases, the signaling sequences in the
database stays at 1 until the installation of the
first module. If several workers are hit for this
database before the first module is installed,
this database registry will be loaded with a signaling
sequence of 1. Since was the default value, any
signal received by workers in this state was ignored
because they thought the registry was previously
not loaded.
Using None to indicate an sequence value
is more accurate and avoids this error
bzr revid: dle@openerp.com-20140116151157-3zlyrg48xqn2lkd0
Fabien Meghazi