See issue #3964 for more detail. Main problem was caused by commit
f0e331e005. It set the key name+'__display' to false when reloading
a record for all field types, but it was only concerned with many2many.
This fix is related to c12a2e1d16
Mutex allow to wait sequentially for mutex
Besides, if one of the onchanges deferreds fails, it does not resolve the mutex until all deferreds of the mutex are resolved.
Besides, we now wait for the onchanges mutex between each field commited value (line 596) in case the commited value leaded to a new onchanges, that we should wait for before commiting the rest of the values.
In multi company environment, the company logo was not updated on company change in the user preferences
This disrupted the end user, as he might think the company change did not happen.
Fixes the issue #1216 (follow the link for more information). The issue
was caused by a hack in list view: the magical suffix __display is used
in render_cell to determine if a many2many field should be updated. This
commit simply makes sure that old many2many fields + __display keys are
cleared.
A better way would be to redesign/refactor the list view to avoid that
hack in the first place. But this would be a much more complex task.
The issue is that when a default searchview is requested, it is
initialized without its view_id. Result: debug mode can't edit the
search view. This commit makes sure that when the field_view_get is
received, the correct view_id is set.
Problem was that when the user types quickly in the search bar and press
enter, the keydown event of the enter key happens before the keypress
event of the last key entered. This means that the autocompletion has
a wrong string. The fix is to move the enter selection detection from
keydown to keyup.
The company_logo is cached and was not refreshed when switching between database, still showing the logo of the company of the previous database
Fixes#2738
This rev. 06104ba553
Added the dirty flag on the o2m field when the editor of the editable list was enabled (meaning that the editable list has been altered)) because the dirty flag was not set correctly by the one2many during the edition, at the time.
It looks like this is now the case
Besides, as now, we valid all the editable list of the form, wether or not the editable list was altered, we must not set the o2m as dirty anymore.
search bar does not suggest date field format based on user's locale and always shows based on mmddyy using Date.parse, opw:615276
Note: starting in 9.0, datejs has been replaced by momentjs, so this
problem should be solved in a better way.
once widget extended with ReinitializeFieldMixin, the event binding with the binary file input and the on_file_change method can be done in initialize_content instead of start
This fix is related to d36c8b5c9b
The add attachment button should be displayed while being in edit mode, but not in view mode
As the widget depends on the form actual mode, the widget should be re-rendered each time the actual mode changes
This is the point of the ReinitializeFieldMixin class
To valid all editable list line, we iterate on the lines and set the editor form with the line value, using set_value.
The _inhibit_on_change_flag should be set to True to avoid triggering on changes events
opw-617395
opw-617036:
In my current timesheet, if you add a a required field on the timesheet ids lines, and add data in the summary tab, it was possible to validate the timesheet lines while requried fields were missing.
When accessing an existing record in form mode directly (enter the url or refresh a page), the daterecord has not been initialized yet. This means that the value of actual_mode will be set to 'edit' before loading the current record (method _actualize_mode() called from do_show()).
This was problematic for one2many fields that we loaded in edit mode, showing add/delete icons/buttons in readonly views. (opw 607910)
This method trigger the focus event when clicking inside the modal, causing a blur event on other widgets resulting in unwanted problem. Typical case is for the jquery datetimepicker where we try to open a select element which will close right after.
Had expected regularity, had not noticed the Python version is really an
ignoble hack where the string "0" (exactly) is interpreted in a specific
manner when used in t-raw and format strings, *and nowhere else*.
Had special-cased the expression everywhere which broke e.g.
<t-set="xxx" t-value="0"/>. Reproduce Python behavior more closely,
and strongly consider swapping out both for something different e.g.
[[content]] or something like that, which doesn't cause problems in
Python but doesn't have to be so specially cased either.
In edit mode, in a *2many with many2many_tags, when adding a new tag when none was set, discarding the form let the new tags displayed while it shouldn't (only a display issue, the tag wasn't added in database)
Fixes#2926
Instead of just hidding it
Otherwise, when opening a new dialog, it is considered that there is a previous opened dialog, and this can lead to refresh the record which opened the hidden/previous dialog
This is related to rev. 57b48602fb
* __content__ can't be used in Python implementation because safe_eval, so use
``0`` from Python implementation instead
* remove postfix from t-call tests because due to implementation details all
whitespace crap following a t-name is added to rendered template in Python
impl, and don't want to normalize whitespace.
* document qweb based (mostly) on JS version
* convert JS qweb tests to (mostly) language-independent XML so they can be
used for JS and Python implementations
* add some more tests (e.g. precedence between t-value and body in t-set)
* remove ``t-import``
* fix parity in foreach(dict) (and rename some variables to make array and
object versions more similar)
When creating a new record in list editable, due to previous commit 6349048, the load_record was called twice and the first record of the current list view (self.dataset.index) was used to fill the new record.
With this, we make sure a new record is indeed created.
Fix the web test to have a default_get call in mock models and increase the number of default_get assertions (for creations in list editable, the default_get is then called twice, not optimal but due to the absence of distinction between empty datarecord and filled with default values).
When reloading a page, the _actualize_mode is called before the record is loaded and the form is displayed partially in edit mode (o2m fields with delete/add icons).
This patch forces to trigger a load_record when the datarecord has not been loaded yet in the do_show merthod (opw 607910)
The previous version of the lib seemed to be in an inconsistent state.
This fix the select2-offscreen class when the page is larger than
10 000px (this situation happens when importing a file with lots
of column).
The cleditor width does not include the margins. Setting 100% will make the editable area too large (104%) on Firefox (opw 611700).
This issue was already fixed in newer releases, from commit 9247c37de7
Nevertheless, it introduced a new issue: while editing the form, the content of widgets html which were located in hidden notebook pages weren't displayed when displaying the page content.
opw-614448
The ManyToOneField widget evaluated the domain before doing a name_search,
but the domain should not be evaluated client side, because it can only
be static at that point. It caused crash in autocompletion because
some variables could not be found in the context (see stage_id in
project.task for example)