Apparently, SEO audit thingies (possibly google's own? unclear) don't
ignore *empty* keyword or description meta tags, then complain that
they're duplicated (there's a use case for SQL-like nulls).
So remove the keyword & description metas entirely if the
corresponding fields are empty.
OPW-815492
Having base_vat installed (but is potentially more general than this)
In the pos frontend, try to save a partner with an incorrect VAT
Before this commit, the generic and cryptic message "your internet is probably down" was displayed to the user
when a python exception occured
After this commit, we display the exception's message
OPW 807656
We should normally stick as close as possible to the version we depend
on in the official DEB packages. This in turn depends on the version of
Debian stable at the time of release - for 9.0 that would be Debian 8
(jessie) and thus Pillow 2.6.1.
However Pillow versions before 3.1.1 suffer from a few issues that could
lead to crashes of Odoo workers. The bugfixes have been backported in
the DEB packages for Pillow, so users of Debian/Ubuntu LTS versions
won't be affected if they are keeing their systems updated.
However it's worth an exception to our rule for pip users.
FWD-PORTS: skip
When the utility method `keep_query()` is used outside an httprequest
context (e.g. rendering a template from a unittest) the following
exception is throwng:
`RuntimeError: object unbound`
The above exception is caused by the method trying to retrieve
parameters from the HTTP request, even if the request object is unbownd.
This change ensures those parameters are retrieved only when exists,
making possible to use it outside the web client
Closes#18841Closes#22942
Closes opw-815322
On latest chromium (as of version 64), accessing rules of a stylesheet
is forbidden for cross-domain.
This commit catch and ignores the error when that happened.
chromium commit: https://chromium.googlesource.com/chromium/src/+/a4ebe08c91
opw-813480
opw-815860
closes#23009
Deleting/replacing an image through summernote gave unexpected results.
As modifiying the related function may be tricky, this commit instead
wrap them to solve the specific erratic behavior: if the range is in/
matches an image, then simply remove the image and move the cursor to
its previous location.
Before this commit, it would do the opposite of what it should:
When 'Do not show this dialog later' is checked, it would remove the
localStorage item instead of creating it.
When unchecked, it would create the localStorage item to prevent the dialog to
shows up again.
Now, we perform the correct behavior.
This closes#10882, closes#11550
Before this commit:
When website default lang was not en_US, the translation of fields* in the
web_editor would not be saved correctly.
Let's illustrate with fr_BE set as main lang:
The web_editor's RPC calls does not send language in the context when editing
a page (6ebcc06), but it does in translation mode.
Note that editing is always performed in the website main lang, fr_BE in this
case.
Therefore:
1. When EDITING:
The backend suppose the language is en_US as there is no lang in the
context (hardcoded everywhere).
It then write in the record itself instead of the ir.translation as it
should since fr_BE is a translation. (In Odoo, every lang is a translation,
en_US is always the main lang)
2. When TRANSLATING (in en_US):
It would also write on model as the web_editor send the language to the
backend when in translate mode. But since en_US is recognized as the main
lang (which is correct), it also writes in the model, overriding the
incorrect fr_BE writing which was also wrote on the model.
This has been fixed in 10.0 with these 2 commits:
036ccbe2955bf2
With these fixes, the lang is correctly added (by the backend) if the website
module is installed.
By ensuring the lang is correctly set, the ORM _write() function will process
the edition and translation datas as it should:
- If it receive en_US, it will write on the model
- If it receive another lang, it will write on ir.translation
This closes#10883, closes#22403
*Not the HTML/XML fields which are handled differently by the ORM/translation
When we validate an order that need to be invoiced, and the customer is
not set, you have to enter a customer, but your order is finalized so
you have a traceback saying it cannot be modified when you try to set
the customer.
When the invoicing fails, we don't set the order as finalized.
A name_search on attribute line would possibly ends up with a domain like:
['|', ('attribute_id', '=', 'm'), ('value_ids', '=', 'm'),
('attribute_id', '=', 'm')]
which would just search the attribute.
This is caused by calling the default name_search. This commit forgo
the default name_search when we are in this attribute line special case.
courtesy-of: @fmdl
opw-813505
closes#22619closes#22760
Name of scripts piped to `odoo shell` were `__builtin__` (`builtins` in
python3), which is not really obvious (and awful if your script must
handle python2 and python3).
Change it to `__main__`, allowing to use to common pattern
`if __name__ == '__main__':`
Before this commit, when connected as a portal user, the user was not
able to post replies to existing replies. This is because the textarea
section was marked to be displayed for connected internal users only.
This commit removes the whole group restriction. Indeed, it makes sense
(as a fix), to show the textarea section for public users too. Indeed,
the "Reply" button is already shown to public users but does not do
anything. With this commit, clicking the "Reply" button shows the
textarea section for everybody and, if the user is not connected,
redirects to the login page when posting the message. This of course
needs usability improvements in master.
Note: the problem had originally been fixed for regular answers with
commit https://github.com/odoo/odoo/commit/b81b03c82d282dc94595a9beb94444041509ef25
Note 2: both mentioned problems were originally introduced by commit
https://github.com/odoo/odoo/commit/9069d0127c176317436b67b23ae5677dd9d53de7
Closes https://github.com/odoo/odoo/issues/22648
When saving a modified mass mailing, the editor will do a number of
things to improve the mail readability accross mail client.
One of those is replacing font awesome icons by image, but firefox acts
differently than other browser. On a display:none iframe, doing
.css('color') or .height() on an element returns respectively
`undefined` and 0.
This caused an error when getting the color that we could solve by
doing a fallback for firefox like this:
window.parent.getComputedStyle($font[0]).color
But to get the height() of an element, it seems we always need the
iframe displayed.
With this change, when the iframe is hidden and the browser is firefox,
the code try to display the iframe (with "visibility:hidden;height:1px")
when this part of the code happen.
note: backport of 10.0 13c326caaa
opw-807180
closes#22701
When clicking on a link in the editor, the "center" alignment button
was always marked as active whatever the real alignment of the link.
This was because it considered the alignment of the text inside the
link instead of the link's one inside its parent.
Sometimes the formatting buttons had no effect or formatted more than
they should (like a whole column when triple clicking on a paragraph
only). This was because of the `listBetween` function implementation.
This is supposed to return all the nodes between one element and another
one... but the logic is wrong without saying from which *point* to start
and which *point* to end. For example, when asking the list between an
element and its parent, the result is very wrong as there is no way to
go from the beginning of an element to the beginning of its parent using
the `walkPoint` function as the `listBetween` function is doing.
Ideally, the function should be entirely fixed but this can be tricky.
For now, the function is just extended to allow to specify points
instead of nodes, and those are used by the `formatBlock` function to
solve the current problem.
... hopefully.
This *must* be forward-ported.
Depending on the browser, the triple-click behavior is different. This
is especially problematic on Chrome, where blank characters at the start
of the neighbor of the clicked element are selected.
A weak attempt to solve that was made by commit https://github.com/odoo/odoo/commit/b85bd358234684974f9af63bc5f28cfabb527767
This commit hopes to solve all the issues once and for all by fully
reimplementing the triple-click for all browsers. The concept is simple:
when a triple-click occurs, select the whole *inner* content of the
deepest DOM *element* that was clicked.
For some unknown reasons, the editor sometimes crashes when dropping a
snippet. As a temporary fix, the non-critical line that causes the
problem is try/catch protected by this commit.
In the case where:
create an invoice with payment terms that contains at least 2 term lines
Validate
create a refund by clicking cancel and reconcile
Before this commit, the refund move lines were wrongly created and reconciled with the invoice's move lines
After this commit, since payment terms are irrelevant on refunds, we set it to false and the problem doesn't raise
OPW 804752
The case: create an invoice of 2000 euros. Payment terms have been configured as:
1) fixed amount 150, 0 days after invoice date.
2) fixed amount 200, 10 days after invoice date.
3) fixed amount 200, 20 days after invoice date.
4) balance, 30 days after invoice date
Before this commit, when issuing a payment of 150 at day 0, that payment got reconciled with the balance line.
After this commit, the payment is reconciled with the right payment term line
OPW 805006
Changing the `deleteContents` method of the editor is tricky. This
commit however dares to add a behavior: if the range does not
represent a selection but only the cursor position and is then asked
to delete its content... it should not do anything as nothing is
selected.
Issue: when the qWeb compress rendered HTML for a better Google PageSpeed
result, some text fields are missing spacebar between fa icons and text,
and between some fields.
The "Mark all read" button either set all messages as read, or if a
domain is given, filter messages on it.
But this filtering is also dependent on not doing sudo so odd
notification to which we would somehow not have any access would never
disappear in the unread number.
Adding a sudo is risky, so this changeset just modify the test on domain
since in most instance the domain is just an empty list (so: []).
The added test before this change would fail at the last assert ("mark
all read should conclude all needactions even inacessible ones").
opw-805185
closes#22277
In 9.0, an "Human Resources / Employee" see only hr.equipment for which
they are set as the "Employee".
But the code automatically subscribed the "Technician" because the field
name was named technically "user_id" which is automatically subscribed
by the system.
This led to a possible issue with read / unread message because of
feature that just happen by coincidence.
For 9.0 only (in 10.0 the field has been renamed technician_user_id and
access rules have been changed).
opw-805185
closes#22273
When computing the total_cost of a landed cost make sure to round the
individual costs of the valuation lines the same way as done by the
ORM (i.e. with the 'Product price' precision).
Otherwise the sum of the additional costs on the valuation lines will
not equal the price of the cost lines. This is wrong and leads to
errors during validation.
opw-804396
- In some languages the kanban images are not displayed because of a bug in human_size method in JS.
This method translates the Unit Size for data and strip with the string with a comma as separator.
When the translatation contains spaces after the commas (like in the French translation), there is extra spaces in the final result.
This then cause an issue with the method is_bin_size that expects a unique space between the value and its unit.
Resulting in a bug in the method 'kanban_image'.
opw #804973#804970#804648
David Arnold