With the new client web from v9, we start the new action before to destroy the old one.
When we switch from Apps (store) to Update and vice-versa, iframe was not loaded correctly.
In case of the Apps widget, we are binding event of the new iframe during the start,
and when it is ready, we destroy the old ifram and so unbind the listener of 'message.apps'.
Using a uniq id, we ensure to bind new event, and only unbind old event during the destroy.
It shouldn't be possible to duplicate ir.model.fields. Add attribute
´duplicate´ which can be set to false to hide the duplication option in the
action menu.
This simple return allows submodules to be able to know when a dialog is shown and modify something in it.
Note from GED: I am aware that this is a IMP in a stable version, and I really don't like that... But it looks like it really helps many people, as shown by the PR, and the risk induced by this commit is definitely extremely low, so I will make an exception.
(PR: #15579)
By default auto_search is set to `True` on window actions, this mean
that for example on a list or kanban view the records will be searched
on the view opening without any user action needed.
Setting it to `False` disable this.
Doing this had two drawbacks:
- depending on race condition, the view could be displayed before the
search view was loaded,
- the code expected `active_search` to be present which was not the case
in this instance.
Before 151c9074 the second issue would not happen (active_search was set
directly resolved if a search was not to be done) and this commit also
wait for the search view being ready before showing the view.
opw-741186
opw-741546
closes#16805
On IE (from 9.0 up to at least IE EDGE 14) we have this behavior
for the method serializeToString of XMLSerializer:
> (new XMLSerializer()).serializeToString($('<b>"</b>')[0])
'<b xmlns="http://www.w3.org/1999/xhtml">"</b>'
> (new XMLSerializer()).serializeToString($('<b>"</b>')[0].firstChild)
'"'
Whilst browser such as chromium or firefox have:
> (new XMLSerializer()).serializeToString($('<b>"</b>')[0])
'<b xmlns="http://www.w3.org/1999/xhtml">"</b>'
> (new XMLSerializer()).serializeToString($('<b>"</b>')[0].firstChild)
'"'
Hence for IE9 and over, if in a `<t t-extend/>` a `t-jquery`
sub-directive (without `t-operation`) is available, we can have
broken javascript if a " is transformed into an " at a unfortunate
location.
This commit favour node.data over XMLSerializer serializeToString to
avoid the possibility of this issue when a text node is processed.
opw-727283
In 9.0 in some particular instances with embedded views, if:
- an x2many list view is editable,
- we go from a form view of an existing record to creating a record,
- the timings of deferred resolving are in a certain order
the x2many list view could be tought as loaded whilst in reality it is
the 'editable list view' row editing 'form view' that has been loaded.
This could cause sometimes issues after bc031379 because the onchange
computation would happen before the list view is really loaded.
This is probably an issue with the deferred system in one of the views,
but it would be rather hard and dangerous to change it.
So this fix introduce an additional test which ignore relational views
if there is an embedded view.
opw-726701
opw-726768
note: only needed for 9.0, the relational views have been changed as of
saas-11 and this issue is not present.
When expanding a textarea of an editable list,
by putting a long description for instnace,
with multiple lines,
some elements were put above the textarea,
making the content partially hidden
opw-709894
Closes#16083
Before this patch, #15920 was happening. The problem was that calling `render_cell` produced a call to [`record.set(column.id + '__display', value)`][1], which triggers the `change` event, which called `render_record` the first time, which called again `render_cell` and produced the 2nd data fetch.
After this patch, `render_record` is only called if there is some place where to put the result, which does not happen in those situations.
There is still the problem that there is one call to name_get for each many2many widget found in a list view (instead of one per full view rendering), but at least they are not two calls!
[1]: https://github.com/odoo/odoo/blob/5d17749ff47c02294d5ff2ae56bbcef9d082562e/addons/web/static/src/js/view_list.js#L1125
The CompoundDomain class allows to regroup several domains with an
implicit "AND"; these domains can be either a string, an array or
another CompoundDomain. A CompoundDomain can then be converted to
an array thanks to pyeval.js.
For example, if a CompoundDomain is initialized with `[A, B]` and
`[C]`, the array conversion gave `[A, B, C]` (which is expected).
However, a hackish method was used with CompoundDomain. If one of the
domain of a CompoundDomain is equal to `["|"]` (an array with the
OR operator in it), the two next subdomains were supposed to be joined
by a OR operator. Indeed, for the case of a CompoundDomain initialized
with `["|"]`, `[A]` and `[B]`, the array conversion gave `["|", A, B]`
(which is expected). However, if initialized with `["|"]`, `[A, B]` and
`[C]`, the array conversion gave `["|", A, B, C]` which is very wrong
as what was expected is `["|", "&", A, B, C]`. The problem is that the
given `[A, B]` contains implicit "&" operators.
This commit fixes the problem by normalizing only if the CompoundDomain
starts with a ["|"] or ["!"] array which is the standard odoo case.
This allows to limit breaking custom code (e.g we want a simple "AND"
of `[A]` and `[B]` to stay `[A, B]`, not become `["&", A, B]`).
The commit also modifies a test so that it checks that the problem is
properly solved.
In a form view, when a field onchange lead to a change on a x2many,
there was two different behavior:
- if the x2many had an embedded view (eg. a tree view inside a form
view) the onchange would notify that it expected the x2many field in
this embedded view to be changed and handled the changes correctly.
- if the x2many had a default view, the onchange ORM would not be
aware the x2many could be modified and would not sent the changes
back causing blank or not updated x2m lines and error on save.
---
Two solutions were birthed to solve the second point:
=> PR #10557 = solving everything
With this PR the onchange in the ORM is aware of every fields in the
current view (even field in a x2m in a x2m in a x2m in a form view) and
if any of these are change the javascript gets back the value of the
fields present in the view.
This PR has currently not been merged by fear of changing too much and
anyway could only be done in master.
=> PR #12249 = if no field for x2many, send its form view fields
With this change, if the ORM onchange is not aware of the fields in the
x2many widget to returns, all the field in the x2m default form view are
returned.
This was merged in bbdf960 but introduced a number of other issue:
- in most situation the x2many is represented by a list view, which may
have fields missing of the form view, so the original is still present.
- the view used may differ from the default form view in other way
(depending on value in context or other possibilities).
- the form view could have fields not present in the form view which
could end up in `write` on fields which should not be written to.
---
This commit reverts bbdf960 and adapts a small part of #10557 so the
x2many with default view works as an embedded x2many. For more than one
level (eg. a x2many in a x2many) this would still not work but it is
only solvable by a PR such as #10557 which could only be targetted for
master.
With this commit:
- the list of fields sent to ORM onchange is computed at the first onchange
- the fields from a x2many field default view is sent for onchange
- the initial onchange on record creation is delayed to when x2many are loaded
closes#12249, closes#15336, closes#15890fixes#11236, fixes#12249, fixes#15129, #15419
opw-705965 opw-716095 opw-715619 opw-710440
note: already present in 10.0 (with a test) as f65475d68
To reproduce:
- Get a slow connection, e.g. activate throttling in debug mode of the
browser.
- Create an invoice with at least one tax.
- Change a price unit.
- Click on 'Save' then repeatedly on 'Validate'
The tax lines are duplicated.
The issue comes from two identical `write` calls whose purpose is to
delete the existing tax line and create a new tax line.
The two calls are issued by the click on 'Save' and the click on
'Validate', which triggers a 'Save' since it doesn't consider the form
as being saved yet.
There were several issues:
- The buttons from the statusbar were not disabled
- The buttons were enabled too soon. They should be enabled again when
the form is reloaded, otherwise the call to 'Validate' doesn't
consider the form saved.
This is an extension of commit
https://github.com/odoo/odoo/commit/bbbf3db581b5cf883
opw-710642
Steps to reproduce:
Go to any pivot view (for example Sales > Report > Sales) and add
the current view to your personnal dashboard (in the search view: Favourites > Add to my Dashboard)
Go to the dashboard, traceback because $buttons was not defined.
opw:707296
Before this commit, when changing measures from the favorites menu (by
selecting a filter with some pivot_measures), the measures in the
dropdown menus were not correctly updated.
This commits make sure that we update the active measures in the menu after
each changes coming from a do_search.
Modern browser implement the autofill specs as described at
https://html.spec.whatwg.org/multipage/forms.html#autofill
This means that there is now a way to tell the browser when a password
field is supposed to be filled with the user's own password (e.g. on a
login field), and when it's supposed to be used to input a new secret
value (e.g. when changing someone else's password).
This can be distinguished by setting the autocomplete attribute
to `current-password` or `new-password`.
In addition, many browsers have decided to entirely disregard the
autocomplete='off' attribute value, because it is felt that using
a secure password manager (as provided by those browsers) is a net
gain for security, allowing users to pick stron passwords.
See also https://developer.mozilla.org/en-US/docs/Web/Security/Securing_your_site/Turning_off_form_autocompletion
Based on this, the framework should generally consider that
password fields in the backend are meant to be used for *setting*
new values (autocomplete="new-password", i.e. no autocompletion!),
whereas the password field on the login page can indeed be autofilled
with the user's current password (autocomplete="current-password").
In addition to making the above changes, this patch also
lets developer specify their own `autocomplete` attribute on
text input fields.
Be a little less strict when not showing o2m list view warning so if the
o2m is not editable (a modal is open instead of editing inline) we still
show warnings.
Also have warnings over modal overlays so they can be used for modals.
opw-702333
The Python method `strftime` raises an error when the date is older than
01-01-1900 ("The datetime strftime() methods require year >= 1900"). If
the user mistyped the date (or was expecting 01-01-16 to be converted
into 01-01-2016), this can lead errors since the ORM uses this method
frequently.
opw-702313
In e0c4ec54ca a change was done to improve the behavior in some
editable list instances, but this introduced a small issue for
list with line creation disabled.
With this change, doing tab or enter at last line should keep the
previous behavior of wrapping around and not creating a new line.
opw-703748
The number of records selected by a FieldCharDomain was only computed
with the base language.
But when selecting the domain with a modal window, or when sending a
mass mailing restricted to a domain, the current user language is used.
With this commit, the number of records is computed with the current
user language.
opw-693836
When a user fills partially a date or datetime field, the system will
somehow autocomplete the value entered. For example, if the user enters
"11//11" then clicks on the Tab key, it will be completed as
"11/11/0000".
However, the year 0000 is not recognized in Python, and this causes a
crash upon saving.
opw-695035
Rev. 3f7d350fe21d9e2afb35a9ce3e2652940e3f259d1 introduced
auto-cancellation of the edit mode when the user clicks
outside the edited line.
The unfortunate downside was the auto-cancellation of the
edit mode also when the user presses ENTER while the
line is not valid.
This is especially annoying when using barcode scanners to
fill in values, as scanners usually send an ENTER keystroke
at the end of the input sequence.
When a many2one field is focused this event will often occur
before the autocompletion popup has a chance to open. At that
point the line is considered invalid, and the edition mode
is cancelled! If no field is considered dirty yet on the line
this simply drops the whole line, otherwise it will open a
navigation confirmation popup. Both results equally prevent
the use of barcode scanners.
This patch makes sure the edit mode is only cancelled when the
user indeed clicked outside the line, and not for other cases
where save() is requested, such as an ENTER keystroke.
If the graph view was left while a nvtooltip was shown (thanks to
keyboard shortcuts or during loading time), the nvtooltip remained
and was only destroyed when leaving the action.
Now they are automatically removed with the view as they are now
placed inside of them instead of inside the body element.
Up to this commit, the JS module system only notified failed modules as
an 'information' instead of an 'error'. This was annoying, since the
runbot takes the result into account, and might be green even though a
module failed to load.
* Tip are never destroyed when change view mode
* unbind methods
* mail: memory leak with unbind methods
* web: memory leak because 'ui-autocomplete' is never destroyed
* web_editor: memory leak for wrong unbind on destroy
* web: memory leak: must remove fileupload handler
* bus: memory leak: correctly debind handlers on window in bus.Bus
* remove handlers on switch buttons and views $buttons
* remove handlers on breadcrumbs
* web: memory leak: correctly debind handlers
* im_odoo_support: memory leak: remove fileupload handler odoo_support
Note: part of this commit was actually authored by aab
Underscore method _.each expects an array like object when a "length"
property is present.
This was an issue with a record having a numeric "length" field set to a
non-negative value. When changing a line the change would not appear on
blur.
This issue was fixed with 0e664c9e9 but introduced back with f0e331e00.
opw-691070
Forward-port of 5d17749ff4 which has been
forgotten during previous forward-port.
When clicking on the trash to remove a line (e.g. a sale order line
in a quotation form view) while editing another line, it produces a
traceback (in most cases), or behaves randomly (less frequently)
like removing both lines, or one of them, or maybe none of them.
Anyway, this feature doesn't work at all, so this commit simply disables
it when we are in edition mode.
In a perfect world, we would fix the feature instead of disabling it,
but the code of the list editable is so instable that it would most
certainly break something else.
Closes#13778
Underscore method _.each expects an array like object when a
"length" property is present.
This was an issue with a record having a numeric "length" field set
to a non-negative value.
opw-691070
Underscore method _.each expects an array like object when a "length"
property is present.
This was an issue with a record having a numeric "length" field set to a
non-negative value. When changing a line the change would not appear on
blur.
This issue was fixed with 0e664c9e9 but introduced back with f0e331e00.
opw-691070
In the widget, the value of reference field is a pair `[val1, val2]`.
When the field is blank, its value is `[false,false]`, thus checking `[false,false] === false`
never succeeds, and this breaks the check for required fields.
This is done by guarding the `on_hashchange` evant handler with the
`clear_uncommitted_changes` function.
Note that the url has already changed when we are in this handler, so we
make sure to restore the old url when the user chose to continue its
edition.
Jeremy Kersten