When computing coverage, the tests are slowed down and the timeout is
often exceeded.
With this commit, HttpCase headless Chrome tests timeout is increased if
coverage is detected.
When Odoo receive a SIGXCPU (CPU time limit reached), it shuts down
immediately. If a headless Chrome is running, it stays alive after the
Odoo shutdown.
With this commit, the signal is intercepeted and the Chrome browser is
properly closed before shutting down the Odoo server.
When Odoo intercept the SIGXCPU signal, an error message is written to
the stderr. Unfortunately, if stderr is redirected to a buffered file, the
message will not appear.
In some situations, Chrome remote debugging is sending an empty list of
opened tabs. In that case, an orphan Chrome process stays alive.
With this commit, Chrome is stopped properly in those situations.
Since rev. 960360af, Date (resp. Datetime) fields return
datetime.date (resp. datetime.datetime) objects. This caused an
issue in the read_progress_bar method which returns a dict, whose
keys are values of the group_by. Indeed, when grouped on a Date or
Datetime field, the key was no longer a string but a datetime.date
or datetime.datetime object, and it crashed when that dict was
turned into a JSON string.
This method isn't tested at all for now, but tests will come
shortly.
Task 1878254.
When switch to list view for tasks, we want them to be grouped
by stage. This was removed at 8169134715 without
justified reason, so this commit reset the previous behavior.
- calendar.event.start and stop are expecting datetime and not date.
Now, if 'All Day' event, start time = 00h00 and stop time = 23h59m59s
- when switching from all day event to not all day event, due to
stop time = 23h59m59s, the rounding for duration leads to make the stop
pass to the next day (00h00). each time you switch between all day and
not all day, the duiration is incremented by 24h.
We remove 1 sec to event stop time to avoid this.
The icon for .o_user_online (e.g. fa-heart for bot) was not displayed as it was
the same color as the one defined in .o_thread_window_header
Use $o-enterprise-primary-color (aka green) to indicate the user is online
Closes#26350
The field SOL on the timesheet list on task should
be visible when the project is billed at employee
rate. This help the user to see which timesheet is
included in which SOL.
The access rights of project.sale.line.employee.map were not
created. This commit provides record rules and accesses for
a project manager to fill the map of a project, and configure
correctly the way of invoicing the timesheet.
With PO's agreements, a such manager should be able to read
services SO and SOL to do that action, without leaking
informations.
Note: reading SO is required as the name_get of SOL fetch
the name of the SO.
Sometimes, we want to give access to sale lines only for some users. But to
allow them to read a SOL, they need the read access of the sale order (as
the name is part of the name_get of SOL), which might break the saleperson
rights. This commit makes the name_get of sOL as to allow fetching
the name of SOL without having the access right on the order.
Now that Date/Datetime fields use python naive datetime, we want
to prevent developper to call or assign aware datetime to
fields. Moreover, if so, the timezone will be stored in the
database, but when reading Datetime fields, we expect naive datetime.
Comparing naive and aware datetime objects leads to errors.
This commit forces developpers to use naive datetime when saving
data in database.
As example, we have to erase the timezone of dates in calendar
modules. We also regroup the 2 assignations in one call to
`write` for performance reason.
Thanks to @dbeguin and @RomainLibert for spending some time
on this bug.
Rev. 279d928693 forced
`server_wide_modules` to include 'base' and 'web', because both contain
controllers that need to be always loaded.
However the patch used a set() that randomized the order of the list,
when combined with Python 3.5's randomized hash function.
This is turn could cause the checksum of asset bundles (web.assets_backend)
to randomly vary, which could cause rapid assets recycling and errors.
Before this commit it was attached to the form, in case your banner was
customized to have the font-color white, the text modal become white on white.
Now we attach to body, to don't have custom design from parent div.
This commit closes opw-1881280
Steps to reproduce the bug:
-Go to Sales > Dashboard > Point of Sale > Settings
-Set nothing as group_by value
-Go back to Dashboard
Bug:
A traceback was raised.
opw:1881444
The resequence mechanism works if the sequence is sorted by its handle field.
Therefore sorting on another field should lock the handle, while sorting on the
handle field should unlock the handle.
opw 1867049
group= should have been groups= so it had no effect.
At the same time the view is already accessible only through a debug menu, so we are fine to always show the list.
Purchase agreements create request for quotation with prefilled field.
The issue is:
date_planned (on RfQ) is filled from schedule_date (on PA). The first
one is a datetime object. The second one is a date object.
This commit makes a right conversion between datetime and date at the
RfQ creation. This issue was introduced by the datetime refactoring
(#24717)
Task id: 1878136
When two transactions conflicts (eg. deleting same data, see [1]) Odoo
will retry the whole transaction several times hoping for the best.
But when rendering template, the error handling would prevent this
feature. For example a recurring issue was:
- loading quickly two times the /web route
- for each recompute the assets
=> this could lead to 2 concurrents transactions that would delete
previous same attachment (ie. DELETE FROM ir_attachment where id=3).
With this changeset, when getting a template fails because of a
transaction rollback, we let the issue bubble up so our retry system is
used.
So instead of a "500 Internal Server Error" page and in log:
bad query: b'DELETE FROM ir_attachment WHERE id IN (3)'
ERROR: could not serialize access due to concurrent update
"GET /web HTTP/1.1" 200 -
"GET /web HTTP/1.1" 500 -
... big traceback ...
load could not load template
we would get the requested page without error and:
bad query: b'DELETE FROM ir_attachment WHERE id IN (3)'
ERROR: could not serialize access due to concurrent update
"GET /web HTTP/1.1" 200 -
SERIALIZATION_FAILURE, retry 1/5 in 0.8920 sec..
"GET /web HTTP/1.1" 200 -
As a side node, the issue was exacerbated in some instances:
- when running a database on another server: assets are recomputed
- when a module was installed/uninstalled: assets may are recomputed
- when updating the source code: assets may be recomputed
- when starting server: requests could be stacked waiting for readiness
- when using google chrome: the "Use a prediction service to load pages
more quickly" option may load a page two times. for example:
-> an URL is entered in address bar
-> a prediction request to it is started URL
-> go to this page (Enter) when that request is not already resolved
-> the prediction request is cancelled and a new request is started
[1] https://www.postgresql.org/docs/9.6/static/transaction-iso.html#XACT-REPEATABLE-READ
note: 10.0 backport of 11.0 #26778
opw-1849167
closes#26785
When two transactions conflicts (eg. deleting same data, see [1]) Odoo
will retry the whole transaction several times hoping for the best.
But when rendering template, the error handling would prevent this
feature. For example a recurring issue was:
- loading quickly two times the /web route
- for each recompute the assets
=> this could lead to 2 concurrents transactions that would delete
previous same attachment (ie. DELETE FROM ir_attachment where id=3).
With this changeset, when getting a template fails because of a
transaction rollback, we let the issue bubble up so our retry system is
used.
So instead of a "500 Internal Server Error" page and in log:
bad query: b'DELETE FROM ir_attachment WHERE id IN (3)'
ERROR: could not serialize access due to concurrent update
"GET /web HTTP/1.1" 200 -
"GET /web HTTP/1.1" 500 -
... big traceback ...
load could not load template
we would get the requested page without error and:
bad query: b'DELETE FROM ir_attachment WHERE id IN (3)'
ERROR: could not serialize access due to concurrent update
"GET /web HTTP/1.1" 200 -
SERIALIZATION_FAILURE, retry 1/5 in 0.8920 sec..
"GET /web HTTP/1.1" 200 -
As a side node, the issue was exacerbated in some instances:
- when running a database on another server: assets are recomputed
- when a module was installed/uninstalled: assets may are recomputed
- when updating the source code: assets may be recomputed
- when starting server: requests could be stacked waiting for readiness
- when using google chrome: the "Use a prediction service to load pages
more quickly" option may load a page two times. for example:
-> an URL is entered in address bar
-> a prediction request to it is started URL
-> go to this page (Enter) when that request is not already resolved
-> the prediction request is cancelled and a new request is started
[1] https://www.postgresql.org/docs/9.6/static/transaction-iso.html#XACT-REPEATABLE-READ
opw-1849167
closes#26778
Steps to reproduce the bug:
- Create a contact without a parent company and set a country on it
- Link this contact to a parent company that has no country defined
- Create an invoice for the contact
- Go to Accounting / Reporting / Management / Invoicing, switch to Pivot view and group by country
=> the invoice for the contact is classified under "Undefined" even though it actually has a
country set when you go see the invoice
opw:1878819
- Create 2 pricelists A and B. Pricelist B relies on A.
- Make B available in the POS, but not A
- Start a POS session
The user faces a black screen without any information on the error.
We add a simple alert so that the user knowns the source of the problem.
opw-1874397
Let's assume a quick create form view with an onchange. Before
this rev., if the user set a value for the char field with
onchange, then clicked on 'Add', the record was created before
the onchange was applied, and thus with incorrect/incomplete data.
The reason is that, when clicking on 'Add', the user triggers a
'change' event on the char field, which notifies the model of its
new value. Then, as 'Add' is clicked, the quick create widgets asks
all widgets to commit their potential (uncommitted yet) changes.
But there is nothing left to commit as the changes have just been
notified. The quick create only waits for this to make the create
RPC, and thus totally ignores the changes that were committed at
the first place.
Task 1878254.
This rev. adds a new function to the mutex concurrency helper:
'getUnlockedef'. This function returns a deferred resolved as soon
as the mutex has no more task to execute (or directly if it is
currently idle).
Such a function was needed for an upcoming fix.
Related to task 1878254.
Before this rev., if the user clicked several times on 'Add' (or
pressed several times 'Enter'), on a KanbanQuickCreate record,
multiple records were created.
This rev. makes sure that this doesn't happen anymore by locking
the widget when it is creating a record ('create' RPC that might
be preceeded by an 'onchange' RPC).
Several test cases are added, to mock the different scenarios that
could occur, as they are all slightly different (depending on if
there is an onchange or not and if the user clicks on Add or press
Enter).
There is still a scenario that needs to be handled, which is the
one with an onchange and where the user clicks on 'Add'. This one
is more tricky and will be handled by a future commit.
Task 1878254.
This rev. removes a deadlock situation that could *theoretically*
occur, but that was actually hard to reach in practice.
Before saving a record (e.g. in a form view), all field widgets
are asked to commit their value (in case they wouldn't have
notified the model yet, e.g. because the user just changed them).
For instance, let's assume that the user fills an input field, and
then triggers the save. In practice, by triggering the save, it
will force the widget to notify the model of its new value (because
it looses the focus). This is also the case using keyboard
navigation, as it actually simulates clicks. So when the widget is
asked to commit its value right after, it has nothing more to
commit.
This is fortunate, because if it actually had something to commit,
we would end up in a deadlock ('commitChanges' is executed in a
mutex, and if the widget triggers a new value, '_applyChange' is
called, and is executed in that mutex as well). In fact, executing
'commitChanges' in a mutex is useless, as '_applyChange', which is
called when a widget has something to commit, is already executed
in the mutex.
So this rev. removes the use of the mutex in 'commitChanges'.
Doing so broke several Many2One tests, because they wrongly relied
on the presence of the mutex for scenarios where the user quick
creates a relational record, and then directly saves (before the
relational record is actually created). We thus handled that
usecase properly to make the tests pass.
Task 1878254.
When an HttpCase browser_js test is started, the screencast is started
and is discarded at the end of the test if no logfile was provided by
the config.
This behavior can impact the performances.
With this commit, the screencast does not start at all if not needed.
A new exception has been added to detect inconsistent reports. However, many are not structured as expected. Thus, the change consists mainly of checking only when the print has to be separated in part to attach it to the respective documents.
We now take the page tag instead of the article, because the article tag is in the layout, so to separate the file, the t-call had to be inside the t-foreach (for the separation) and so possibly read the assets several times. Since all templates have the first tag (in the t-foreach) tag page, the choice is more sensible.
The javascripts files used in readonly were taken out of the editing assets.
A turn was made but not activated, it works locally, but on the runbot, an error is triggered when generating assets after choosing the settings.
The default value for failure_type was the 'NONE', instead of being
simply False/NULL. This represents a significant waste of tablespace
with no added value.
Switching to true NULL value for representing the absence of error
also save some precious time during database upgrades on databases with
large numbers of emails.
While fixing the values for failure_type, also fixed a few typos and
untranslated strings.
See original feature at 133eeb1bbf
This commit changes two things in the time range menu
appearance:
- the order of range options
- the spacing between them (now two different groups
are separated by an empty lines)
Some tests had to be changed accordingly.
This commit fixes a bad behavior of the pivot view. By the
past, if there were no data to present in the pivot view, no
information on the pivot view configuration was returned by
the get function. This does not cause problems to the pivot
renderer (by design) but causes problems elsewhere, for
instance in a dashboard view that contains a pivot view.
Indeed, at each update, the dashboard view reinstantiates its
subviews, so that information can be lost if a subview is of
pivot type.
The _mergeData function used in pivot view comparison mode
was not designed to accept the absence of the __count key or
null values for measures in the objects returned by a 'read_group'.
The absence of the __count key would result in a traceback and null
values for measures into wrong calculations (infinities).
One could ask why null values did not cause problem before.
The simple reason is that the formatters used to format cell values
render a 'null' value like they render a zero.
This fix makes explicit that null values have to be considered
as zeros and avoid in that way Infinity values when computing
variations.
The search view in mobile mode has been redesigned and looks a bit different
from the search view in desktop mode. This commit adapts the time range menu
appearance in mobile mode.
Exporting data of a pivot view in comparison mode ends in a
crash. The root cause is that _t was replaced by _lt
so that some strings were replaced by objects with a toString method.
This does not cause problems when rendering a pivot view since
both QWeb or JQuery call that method. The function _exportData does not.
This commit fixes that situation and adds a test of _exportData.
The buttons 'Flip Axis', 'Expand All', and 'Download xls' are
now deactivated when no data are displayed in the pivot view
(those buttons are useless in that situation).
Christophe Monniez