When the content of a Many2Many tag is too large, it is truncated. This
is an issue since there is no easy way to see it.
The fix adds a title on the tag with the content.
X-port of commit odoo/enterprise@bd3bfe0422
The action object, and in particular its context, is modified by the Action
Manager and by some views (e.g. sales_team_dashboard Kanban). So, as the action
is stored in cache, the same object was returned each time the action was
requested, and previous modifications of its context still applied.
This produced a bug with the sales_team_dashboard when following those steps:
- click on the link to today's activities, it opens an action with
'search_default_today' in the context,
- come back to the dashboard
- click on the link to this week's activities, it opens an action with
'search_default_this_week' in the context, but the formerly added key
'search_default_today' is still there as well, so the domain is wrong.
When the user visited a view for the second time, the extra actions in the sidebar
were broken. The reason is that the sidebar code modified in place the
fields_view information.
Note: This was not a problem before, because the field_view_get call was done each
time the view was loaded, but now it is cached.
M2O use the autocomplete lib to display its content. The problem is
that if a result is empty (or only composed of whitespaces), a
divider is displayed in the dropdown.
Commit 1d9f7d279a was fixing this by only displaying non-empty
name results... but it did not work as many models implement the
name_search method the wrong way.
The new solution is to add a default display for record values which
are empty strings, at the right places in the webclient components.
When we download a file (as can be done with attachments in Odoo),
firefox is opening the link in current window but cancel the action once
it is aware the resource is to be downloaded.
But this get firefox into a corrupt state for some contenteditable
features, for example document.queryCommandState which is needed for the
editor will after downloading a file throw an error consistently instead
of working as expected.
It can be reproduced on firefox (at least from version 32 to 46) with
the following snippet:
https://jsfiddle.net/s246u3ay/3/
The issue is probably linked to the following reported issue since 2005:
https://bugzilla.mozilla.org/show_bug.cgi?id=297494
This commit avoid the bug by opening attachments download links in a new
window instead of the current one.
closes#12109
opw-676918
A fonts.less file which contains the definition of Lato was created
in saas-11 and placed in the backend assets. It should have replaced
the lato/stylesheet.css in the common assets.
Also adapt point_of_sale assets to use the same file.
Form views with a FieldBinaryImage produced a traceback (only visible in the
server-side logs) when creating a new record from an existing one, because the
fields were re-rendered before being updated (in this case reset to their
defaults). In the case of the FieldBinaryImage, as the field wasn't reset yet,
we tried to load the image of a null id, which produced the traceback.
This rev. ensures that the fields are updated before being re-rendered when
switching to another (possibly new) record.
Note: this is spaghetti code, several re-rendering are triggered when switching
between records (e.g., when updating the fields value, when actualizing form
view mode). This should definitely be refactored/simplified.
Before this rev., if an error occured during an RPC handled by the
data_manager, the RPC result was stored in cache anyway, and there was thus no
other way than refreshing the page to get correct result for that RPC.
The left margin of the graph doesn't take into account the width of the
maximum Y-axis value. In case the values are high (> 10^7), they are
cropped.
The fix is to adapt the left margin as a function of the size of the
maximum value.
opw-676538
Before this commit, in some cases, clicking on 'save' button, then on
other buttons (like 'validate' in invoices) had the unfortunate effect
of applying twice the changes in some cases.
More specifically the problem appears when:
1. an onchange is triggered (for example, taxes in invoices)
2. the user clicks on 'save', buttons are disabled
3. the 'save' actions is completed, the buttons are enabled, but
the on_button_save still does some work: triggering event, and
reloading the view
4. the user clicks on 'validate'
5. the onchange completes, this triggers some more work
Our only protection against such interactions is right now the fact that
buttons are disabled/enabled. In my opinion, there should be a way
to guarantee that actions are properly serialized (that is actually what
a mutex is for... just saying). But the code of the form view/list view is
too coupled, and such changes are quite dangerous. A real refactoring
of those is needed at some point in the future.
When clicking on a m2o without typing any search text, the dropdown
shows results for the empty search string. The problem is that, on
some databases, records with empty names exist. However ui-autocomplete
library shows empty string label as dropdown divider.
Since a task is in development to prevent users from having non
trimmed record names, the solution that was chosen here is to
exclude empty names from the results. As the name_search can't evaluate
trimmed field, this commit fix the problem for up to three spaces
record names.
When the user searches for date (or a datetime) in the search bar, the
suggested date can become completely wrong. For example, in the French
localization where the date format is DD/MM/YYYY, the system will search
for the date MM/DD/YYYY. This is even worse when the date is not fully
written: DD/MM could be converted to MM/DD/0000, which doesn't make any
sense. This is related to the moment issue
https://github.com/moment/moment/issues/1407.
The fix is to make sure that the input date has a correct format thanks
to `parse_value`. The latter converts the date into a fixed format.
Fixes#11876
opw-676604
The goal of this commit is to have better search results when fuzzy
searching. Before this commit, the position of a match was irrelevant
to the search results. After, a match in a first position has more
weight than a match in last position. This is better, because we often
want to find results using the first letters.
This is controversial, but I could not find a better way to do this
without rewriting the whole library myself (which might not be a bad
idea, it is only 50 lines long)
The current behavior was allowing going on the previous row if the
user is at the start of an editable element and on the next row if the
user is at the end of an editable element. The problem was, we did not
want this behavior when the editable element is a select or m2o (as
the up and down input should open the select/m2o dropdown instead).
By putting the shortcuts lowercase, it is easier to remember and to user.
Instead of Alt + Shift + letter, we have Alt + letter for Windows and Linus users
Instead of Control + Alt + Shift + letter, we have Control + Alt + letter for mac users
The combinations Alt + d and Alt + e are reserved to the browsers, so other letters are chosen
for these actions
When opening a record from a list view, the Edit button in the control panel
sometimes bounced. This was because we listened to event 'mouseup', and
astonishingly, the 'mouseup' event of the click in the list view was catched
by the form view.
The reason why we listened to 'mouseup' instead of 'click' is that some
widgets need the event to be propagated to the body. So this fix simply makes
the form view listen to 'click' instead of 'mouseup', but removes the call to
stopPropagation() on the catched event.
With this commit, actions can have a type=fullscreen. This means that
the navbar will be hidden, for the action and for all subsequent actions
open in the breadcrumb
There is no real security behind this, this is just a feature for
convenience, so users can bookmark an url an open some action in a way
that they are not distracted by anything. In short, this is some kind of
light 'kiosk' mode.
Nicolas Martinelli