Steps to reproduce
==================

- Use a mobile viewport
- Go to planning
- Click on an empty cell
- Click on the date
- On the bottomsheet, click on the time at the bottom
- Try to type 12:34

=> Only 1 is entered and then the start_datetime field is focused behind the bottom sheet

Cause of the issue
==================

The <input type="time"/> listen to the onchange event. We listen to rawPickerProps changes using a reactive call. shouldFocus is then set to true, and the focus is done after the next render.

The onchange event is called at a different time depending on the platform.

On IOS and Firefox desktop: after changing hours or minutes
On Android: once the apply button is clicked
On Chrome desktop: After entering a single char

Solution
========

The bottomsheet is only displayed when env.isSmall && hasTouch(). It doesn't make sense to focus the input, since we don't handle the keyboard in that case.

opw-6386252

Forward-Port-Of: odoo/odoo#283782
Forward-Port-Of: odoo/odoo#276617

### Issue:
When multiple partners share the same email address, `_mail_find_partner_from_emails` may resolve to the wrong partner when the input is a formatted email like "`Name <email>`"

This affects use cases like email templates using `{{object.company_id.email_formatted}}` as sender, where the wrong company partner could be selected

### Cause:
The lookup in `done_partners` only matched on `email_normalized`, which cannot distinguish partners sharing the same email but with different names

The `email_formatted` field carries both name and email, allowing an exact match when the input is a formatted email

### Steps to reproduce:
- Install `account`
- Create an Email Template (Applies to: account.move, From: {{object.company_id.email_formatted}})
- Create a second company B with the same email as the default (e.g. info@yourcompany.com)
- In Settings (logged in as company B), set a Fiscal Position (e.g. US Taxable)
- Create an Invoice on company B
- In the chatter, click Send message, click the expand arrows button, use the three dots menu to select the template
- Send and check the Sender in the chatter

Before the fix, the sender resolves to the default company even though the invoice belongs to company B

opw-6260992

Forward-Port-Of: odoo/odoo#283875
Forward-Port-Of: odoo/odoo#269509

Steps to reproduce:
- Enable presets and set a preset carrying a tax-replacing fiscal position (e.g. "Takeout") as the default preset of the POS config
- Take an order with another preset that has no fiscal position (e.g. "Dine In")
- Refund that order from the ticket screen

Issue:
The refund is taxed with the fiscal position of the default preset instead of the one of the refunded order. The ticket screen displays the refund lines with the taxes of the original order, but the refund that is actually created maps them through the wrong fiscal position. With tax-included prices the totals still match on screen, so the operator only sees the discrepancy after going back.

Cause:
The destination order of a refund is an empty order, created with the default preset and therefore with that preset's fiscal position. In TicketScreen.onDoRefund, the fiscal position of the refunded order was only copied onto it when the refunded order had one, so an order taken without a fiscal position kept the default preset's one. When an already existing empty order is reused as destination, whatever fiscal position was last set on it survives for the same reason.

Fix:
Always assign the fiscal position of the refunded order to the destination order, an empty one included, so a refund is taxed exactly like the order it refunds instead of silently switching. The preset itself is left untouched: it drives the ordering workflow (timing slot, customer identification) which must not be imposed on a refund.

opw-6442664

---
I confirm I have signed the CLA and read the PR guidelines at www.odoo.com/submit-pr

Forward-Port-Of: odoo/odoo#282560
Forward-Port-Of: odoo/odoo#280672

Steps to produce:
---
- Install the `Sales` module.
- In Settings, enable `Automatic Invoice`.
- Also enable the Demo payment provider.
- Create a sale order with a total of `800` and confirm it.
- Generate a payment link for `200` from the gear icon and pay it.
- Generate a second payment link for `300` and pay it.
- Generate a final payment link for the remaining `300` and pay it.

Issue:
---
- After the first payment (200), an `invoice of 200` is created. Correct.
- After the second payment (300), an` invoice of 500` is created instead of 300.
- After the third payment (300), an `invoice of 100` is created instead of 300.

Root cause:
---
- The down payment invoice uses `order.amount_paid`, the cumulative sum of all transactions on the order, instead of the amount of the latest payment. This causes invoices to be sized off the running total instead of the individual payment delta.

Fix:
---
- Compute the invoice amount as `order.amount_paid - order.amount_invoiced` (the unpaid) instead of passing the cumulative `amount_paid` directly.

opw-6324036
---
I confirm I have signed the CLA and read the PR guidelines at www.odoo.com/submit-pr

Forward-Port-Of: odoo/odoo#283492
Forward-Port-Of: odoo/odoo#273099

Before this commit, clicking "Close all conversations" from the chat hub as an operator with several live chat windows open (fresh database with im_livechat demo data, logged in as admin) sometimes crashed:

    TypeError: Cannot read properties of undefined (reading
    'findIndex')
        at Proxy.close (mail/static/src/core/common/chat_window_model)

or with "RangeError: Maximum call stack size exceeded" in the record machinery. The overflow interrupts that machinery anywhere, hence the varying crash.

This happens because closing all windows leaves every live chat channel at once, and the resulting bus payloads reach the store out of their transaction order. Resolving an out-of-order many command derives a REPLACE from the field's history, and applying it updates the inverse field of each removed record. The problem is that these inverse updates are resolved against the same out-of-order revision:

- on the many side, they derive another REPLACE on the field they come from, re-entering the resolution before the field is updated, until the stack overflows;
- when their revision is older than the field's last replace, they are dropped, leaving the two sides of the relation out of sync.

Reminder that inverse updates go through `updateFields()` since "[FIX] mail: no infinite loop on discuss page load", so that the version history records them.

This commit fixes the issue by always applying client-generated updates (inverse echoes, computes, direct field writes): they reflect state the client already changed elsewhere, so only server data is subject to version resolution. They are still recorded in the history.

task-6450249

Forward-Port-Of: odoo/odoo#283204

When updating mail notifications during `mail.mail._send()`,

a `SerializationFailure` raised while flushing the notification recordset leaves the current transaction in an aborted state.

As `_send()` continues handling the exception, accessing fields:
 - https://github.com/odoo/odoo/blob/127f1316540ec6cc6880ea3515e6923ba3903bc7/addons/mail/models/mail_mail.py#L816

So, any subsequent SQL query fails with

`InFailedSqlTransaction`, masking the original concurrency error.

Avoid accesing to `mail.message_id` with aborted cursor, preserving the original `SerializationFailure`.

A regression test is added to simulate a concurrency failure during

`flush_recordset()` and verify that the cursor is no longer used dirty

The logger for the unittest without the fix is the following:

```log
During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "odoo/addons/mail/models/mail_mail.py", line 719, in _send
    notifs.flush_recordset(['notification_status', 'failure_type', 'failure_reason'])
  File "<string>", line 3, in flush_recordset
  File "unittest/mock.py", line 1139, in __call__
    return self._mock_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "unittest/mock.py", line 1143, in _mock_call
    return self._execute_mock_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "unittest/mock.py", line 1204, in _execute_mock_call
    result = effect(*args, **kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^
  File "odoo/addons/test_mail/tests/test_message_post_concurrent.py", line 93, in mocked_mail_notification_flush_recordset
    return original_flush_recordset(self, *vals, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "odoo/odoo/models.py", line 6788, in flush_recordset
    self._flush(fnames)
  File "odoo/odoo/models.py", line 6852, in _flush
    model.browse(some_ids)._write_multi(vals_list)
  File "odoo/odoo/models.py", line 4938, in _write_multi
    self.env.execute_query(SQL(
  File "odoo/odoo/api.py", line 993, in execute_query
    self.cr.execute(query)
  File "odoo/odoo/sql_db.py", line 371, in execute
    res = self._obj.execute(query, params)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
psycopg2.errors.SerializationFailure: could not serialize access due to concurrent update

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "odoo/addons/test_mail/tests/test_message_post_concurrent.py", line 107, in test_mail_send_dirty_cursor
    mails.send()
  File "odoo/addons/mail/models/mail_mail.py", line 652, in send
    self.browse(batch_ids)._send(
  File "odoo/addons/mail/models/mail_mail.py", line 818, in _send
    mail.id, mail.message_id)
             ^^^^^^^^^^^^^^^
  File "odoo/odoo/fields.py", line 1309, in __get__
    self.compute_value(recs)
  File "odoo/odoo/fields.py", line 1491, in compute_value
    records._compute_field_value(self)
  File "odoo/odoo/models.py", line 5302, in _compute_field_value
    fields.determine(field.compute, self)
  File "odoo/odoo/fields.py", line 113, in determine
    return needle(records, *args)
           ^^^^^^^^^^^^^^^^^^^^^^
  File "odoo/odoo/fields.py", line 710, in _compute_related
    record[self.name] = self._process_related(value[self.related_field.name], record.env)
                                              ~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^
  File "odoo/odoo/models.py", line 7083, in __getitem__
    return self._fields[key].__get__(self)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "odoo/odoo/fields.py", line 1272, in __get__
    recs._fetch_field(self)
  File "odoo/odoo/models.py", line 4120, in _fetch_field
    self.fetch(fnames)
  File "odoo/addons/mail/models/mail_message.py", line 756, in fetch
    return super().fetch(field_names)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "odoo/odoo/models.py", line 4158, in fetch
    fetched = self._fetch_query(query, fields_to_fetch)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "odoo/odoo/models.py", line 4245, in _fetch_query
    rows = self.env.execute_query(query.select(*sql_terms))
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "odoo/odoo/api.py", line 993, in execute_query
    self.cr.execute(query)
  File "odoo/odoo/sql_db.py", line 371, in execute
    res = self._obj.execute(query, params)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
psycopg2.errors.InFailedSqlTransaction: current transaction is aborted, commands ignored until end of transaction block
```

Real error in production:

```log
2023-04-15 10:43:14,978 106451 ERROR my_db odoo.sql_db: bad query: UPDATE "mail_notification" SET "failure_reason" = "__tmp"."failure_reason"::text, "failure_type" = "__tmp"."failure_type"::VARCHAR, "notification_status" = "__tmp"."notification_status"::VARCHAR FROM (VALUES (4426629, 'Error without exception. Probably due to concurrent access update of notification records. Please see with an administrator.', 'unknown', 'exception')) AS "__tmp"("id", "failure_reason", "failure_type", "notification_status") WHERE "mail_notification"."id" = "__tmp"."id" ERROR: could not serialize access due to concurrent update
```

```log
2023-04-14 10:43:14,978 106451 ERROR my_db odoo.sql_db: bad query: UPDATE "mail_mail" SET "failure_reason"='Error without exception. Probably due do sending an email without computed recipients.',"headers"='{''X-SMTPAPI'': ''{"ip_pool": "Transactional"}'', ''X-Odoo-Objects'': ''sale.order-1436960''}',"state"='exception',"write_uid"=1,"write_date"=(now() at time zone 'UTC') WHERE id IN (2548540)
ERROR: current transaction is aborted, commands ignored until end of transaction block
```

Description of the issue/feature this PR addresses:

Current behavior before PR:

Desired behavior after PR is merged:


# UPDATE 2026-07-22

The reviewer requested to remove the large docstring

For record, the docstring was

```python
    """Reproduces a concurrency scenario where `mail_mail._send()` fails with a PSQL SerializationFailure after
        flushing `mail.notification` records. After such a failure, the cursor is left in an aborted
        (`InFailedSqlTransaction`) state, so any further SQL access (e.g. reading `mail.message_id` like
        https://github.com/odoo/odoo/blob/127f1316540ec6cc68/addons/mail/models/mail_mail.py#L816)
        would raise a new error masking the original SerializationFailure.

        Setup:
        - Uses a separate `cursor()` to create and commit a message with its `mail.mail` and `mail.notification`
        records, so they are visible to a second, concurrent transaction.

        Concurrency simulation:
        - `MailNotification.flush_recordset` is patched so that, right before the real flush runs, a second cursor
        updates the same `mail.notification` records (`failure_reason`). This forces PSQL to raise a
        SerializationFailure when the original transaction tries to flush those rows.

        Assertions:
        - `SerializationFailure` is raised confirming the concurrency conflict.
        - `mail_mail._send()` logs the expected error message containing the mail `id` and `message-id`

        Cleanup: created records are unlinked in `finally`
        """
```

# UPDATE 2026-07-23

The reviewer requested to remove the unittest

For record, the unittest was

```diff
diff --git a/addons/test_mail/tests/test_message_post.py b/addons/test_mail/tests/test_message_post.py
index 53dd5b9eec52..46a3958a5bff 100644
--- a/addons/test_mail/tests/test_message_post.py
+++ b/addons/test_mail/tests/test_message_post.py
@@ -7,17 +7,21 @@ from datetime import datetime, timedelta
 from freezegun import freeze_time
 from itertools import product
 from markupsafe import escape, Markup
+from psycopg2.errorcodes import SERIALIZATION_FAILURE as SERIALIZATION_FAILURE_CODE
+from psycopg2.errors import SerializationFailure
 from unittest.mock import patch
 
-from odoo import tools
+from odoo import SUPERUSER_ID, api, tools
 from odoo.addons.base.tests.test_ir_cron import CronMixinCase
-from odoo.addons.mail.tests.common import mail_new_test_user, MailCommon
+from odoo.addons.mail.models.mail_notification import MailNotification
+from odoo.addons.mail.tests.common import mail_new_test_user, MailCommon, MockEmail
 from odoo.addons.test_mail.data.test_mail_data import MAIL_TEMPLATE_PLAINTEXT
 from odoo.addons.test_mail.models.test_mail_models import MailTestSimple
 from odoo.addons.test_mail.tests.common import TestRecipients
 from odoo.api import call_kw
 from odoo.exceptions import AccessError
-from odoo.tests import tagged
+from odoo.modules.registry import Registry
+from odoo.tests import TransactionCase, get_db_name, tagged
 from odoo.tools import mute_logger, formataddr
 from odoo.tests.common import users
 
@@ -2244,3 +2248,49 @@ class TestMessagePostLang(MailCommon, TestRecipients):
                                     self.assertIn('html lang="es_ES"', email['body'])
                                 else:
                                     self.assertIn('html lang="en_US"', email['body'])
+
+
+@tagged('database_breaking')
+class TestMessagePostConcurrent(MockEmail, TransactionCase):
+    """Mail concurrency edge cases that require real, separately committed transactions
+    instead of the usual rollback-based TransactionCase isolation.
+    """
+
+    def test_mail_send_dirty_cursor(self):
+        """Reproduces SerializationFailure `mail_mail._send()` fails,
+        the cursor is left in an aborted state, so any further SQL access would raise a new error
+        (e.g. reading `mail.message_id` like
+        https://github.com/odoo/odoo/blob/127f1316540ec6cc68/addons/mail/models/mail_mail.py#L816)
+        """
+        original_flush_recordset = MailNotification.flush_recordset
+
+        def mocked_mail_notification_flush_recordset(self, *args, **kwargs):
+            with Registry(get_db_name()).cursor() as cr:
+                cr.execute('UPDATE mail_notification SET failure_reason = %s WHERE id IN %s', ('Forced Concurrent Update', tuple(self.ids)))
+            return original_flush_recordset(self, *args, **kwargs)
+
+        recs2unlink = []
+        with Registry(get_db_name()).cursor() as cr:
+            env = api.Environment(cr, SUPERUSER_ID, {})
+            partner = env.ref('base.user_admin').partner_id
+            try:
+                message = partner.message_post(body='Hello', message_type='comment', partner_ids=[partner.id], mail_auto_delete=False, force_send=False)
+                notifs = env['mail.notification'].search([('notification_type', '=', 'email'), ('mail_mail_id', 'in', message.mail_ids.ids)])
+                self.assertTrue(notifs)
+                mails = message.mail_ids
+                recs2unlink.extend([notifs, mails, message])
+                cr.commit()
+
+                mails = self.env[mails._name].browse(mails.ids)
+                with (
+                    mute_logger('odoo.sql_db'), self.assertRaises(SerializationFailure) as exc, self.mock_mail_gateway(),
+                    patch(f'{MailNotification.__module__}.{MailNotification.__name__}.flush_recordset', autospec=True, side_effect=mocked_mail_notification_flush_recordset),
+                    self.assertLogs('odoo.addons.mail.models.mail_mail', level='ERROR') as log_capture,
+                ):
+                    mails.send()
+            finally:
+                for rec2unlink in recs2unlink:
+                    env[rec2unlink._name].browse(rec2unlink.ids).unlink()
+
+            self.assertEqual(exc.exception.pgcode, SERIALIZATION_FAILURE_CODE)
+            self.assertIn(f'Exception while processing mail with ID {mails.id} and Msg-Id \'{mails.message_id}\'.', [record.message for record in log_capture.records])

```

---
I confirm I have signed the CLA and read the PR guidelines at www.odoo.com/submit-pr

Forward-Port-Of: odoo/odoo#279897
Forward-Port-Of: odoo/odoo#274089

Problem: 
When a `table` with a `color`/`backgroundColor` contains a nested `table`, `distributeTableColorsToAllCells` propagates the outer table's color to every `td` in the subtree, including cells belonging to the inner table. The inner table's own color is then discarded since its `td`s already have a value.

Cause: 
`table.querySelectorAll("td")` returns every `td` in the entire subtree, not just the table's own direct cells.

Solution:
Scope the selected `td`s to `td.closest("table") === table`, so a table's color is only distributed to its own cells.

Steps to reproduce:
1. Add a `background-color` to an outer `table`.
2. Nest a `table` with a different `background-color` inside one of its cells.
3. Load/normalize the content in the editor.
4. Observe both tables' cells carry the outer table's color.

opw-6438972

---
I confirm I have signed the CLA and read the PR guidelines at www.odoo.com/submit-pr

Forward-Port-Of: odoo/odoo#283373
Forward-Port-Of: odoo/odoo#281413

**Steps to reproduce:**

-  Install the Serbian EDI module `l10n_rs_edi`.
-  Configure eFaktura credentials on the company.
-  Create and confirm a Serbian customer invoice.
-  Send the invoice to eFaktura.

**Observed Behavior:**

When the eFaktura API returns an HTTP error, Odoo only displays the generic exception generated by `requests`, for example an HTTP 400/500 error.

The actual error information returned by eFaktura in the response body is not shown to the user, making it difficult to understand why the invoice was rejected.

**Cause:**

`_l10n_rs_edi_send` catches `HTTPError`, `Timeout`, and `ConnectionError`, but the error message is built only from the Python exception.

For HTTP errors, the eFaktura API may return a response containing more precise information such as:

```json
{
    ErrorCode: ...,
    Message: ...
}
```

This response was not being used when displaying the error in Odoo.

**Fix:**

When an HTTP response is available and contains an eFaktura error payload, use the returned `ErrorCode` and `Message` as the error displayed on the invoice. Fallback to the existing connection/HTTP exception message when no usable API response is available.

opw - 6453653

Forward-Port-Of: odoo/odoo#281490

Steps to reproduce
------------------
1. Configure a product with a subcontracted BoM and a subcontractor.
2. Create a purchase order of 10 units for that product and confirm it.
3. Receive the 10 units.
4. On the receipt, use "Return for Exchange" on 3 units and validate both the return and the exchange receipt.

Issue
-----
After the exchange, the 3 units stay in the subcontracting location instead of reaching `WH/Stock`, and the received quantity on the purchase order line stays at 7 instead of 10.

`mrp_subcontracting` overrides `_prepare_move_default_values` to force the move `location_dest_id` to the subcontractor location for every `is_subcontract` move:
https://github.com/odoo/odoo/blob/d9c06a66356dd9d5a50821b8cde6194967353c18/addons/mrp_subcontracting/wizard/stock_picking_return.py#L20-L25
That is correct for the return, but the same override also runs for the exchange re-receipt, an `incoming` picking whose destination should be the stock location from `return_type.default_location_dest_id`:
https://github.com/odoo/odoo/blob/d9c06a66356dd9d5a50821b8cde6194967353c18/addons/stock/wizard/stock_picking_return.py#L137-L153

The exchange move then goes from the subcontracting location back to itself, so validating it nets zero and `WH/Stock` never receives the units. Skipping the override when `new_picking.picking_type_id.code` is `incoming` lets the exchange land in stock.

The received quantity must also count that receipt. `_should_count_for_quantity_received` only counts `supplier` or `transit` sources:
https://github.com/odoo/odoo/blob/d9c06a66356dd9d5a50821b8cde6194967353c18/addons/stock/models/stock_move.py#L330-L331 so the exchange, sourced from the internal subcontracting location,
is skipped while the return still subtracts its quantity. Counting subcontracting-sourced moves:
https://github.com/odoo/odoo/blob/d9c06a66356dd9d5a50821b8cde6194967353c18/addons/mrp_subcontracting/models/stock_move.py#L312-L314 restores `qty_received` to 10.

opw-6410978

Forward-Port-Of: odoo/odoo#282666
Forward-Port-Of: odoo/odoo#279431

### Steps to Reproduce:
1. Have a product where Track Inventory is enabled and the product category is FIFO and Perpetual
2. Create a PO for the product
3. Validate the receipt
4. Update the quantity on the PO to 0
5. Create Accrued Expense Entry > Traceback

### Description of the issue/feature this PR addresses: 
**Issue:** Currently when generating an Accrued Expense Entry for a PO where quantity on the line is updated to 0, the system crashes with an RPC error. This happens because reducing the line quantity to 0 sets the overall order amount to 0.0. Then. when the accrued orders wizard tries to calculate line-item ratios, it triggers a `ZeroDivisionError`.

**Solution:** We can add a zero-check fallback condition when computing the line ratio inside `_compute_move_vals` in the
`AccountAccruedOrdersWizard` class. The ratio calculation now defaults to 0.0 if the order total is zero, preventing division by zero.

### Current behavior before PR: 
Triggering the Accrued Expense Entry wizard on a PO with a changed quantity of 0.0 causes a `ZeroDivisionError` server error. The user receives an RPC error dialog and cannot proceed with creating the journal entry.

### Desired behavior after PR: 
The wizard should be able to process Purchase Orders with a line quantity of 0 without throwing an RPC error. The system should now cleanly generate the accrual entry based on received quantities.

opw-6459403

---
I confirm I have signed the CLA and read the PR guidelines at www.odoo.com/submit-pr

Forward-Port-Of: odoo/odoo#281808

Currently, an error occurs when a user opens the calendar.

Steps to Reproduce:
 - Install the `hr_calendar` module.
 - Go to `Employees` and create an `employee`.
 - Under the `Settings tab`, set the employee's `user` to `Administrator`.
 - Create a `new company` and switch to it.
 - Open the `Calendar` and set the `scale` to `Month`.

`TypeError: reduce() of empty iterable with no initial value`

When the user opens the calendar, it fetches the unusual days for the selected attendees. 
By default, the current user's partner is set as an attendee [1]. It then computes the 
schedule for the attendee [2], which returns an empty dictionary [3] because the linked 
employee belongs to a different company than the current company, as restricted by the 
domain [4]. This empty dictionary is then passed to reduce() to intersect the schedules, 
which raises an error because the iterable is empty.

This commit ensures that when no schedule is found for the attendees, an empty set is returned.

[1]: https://github.com/odoo/odoo/blob/658018684d781fef8bf77a77f1e050d1eb16937c/addons/hr_calendar/models/calendar_event.py#L38-L40

[2]: https://github.com/odoo/odoo/blob/658018684d781fef8bf77a77f1e050d1eb16937c/addons/hr_calendar/models/res_partner.py#L123-L130

[3]- https://github.com/odoo/odoo/blob/658018684d781fef8bf77a77f1e050d1eb16937c/addons/hr_calendar/models/res_partner.py#L40-L42

[4]- https://github.com/odoo/odoo/blob/658018684d781fef8bf77a77f1e050d1eb16937c/addons/hr_calendar/models/res_partner.py#L19-L25

sentry-7672490836

---
I confirm I have signed the CLA and read the PR guidelines at www.odoo.com/submit-pr

Forward-Port-Of: odoo/odoo#282866

Issue before this commit:
========================
Currently, when the reception report option is enabled for the operation type under Print on Validation, validating a transfer with reserved quantities for the next transfer automatically tries to print the reception report. However, the print fails with `External ID not found in the system: stock.stock_reception_report_action`, preventing the transfer from being validated.

Steps to Reproduce:
=========================
- Install the stock module.
- In Inventory → Configuration → Operation Types → Receipts, enable Reception Report under Print on Validation.
- Create a delivery and receipt for Product A, both with a quantity of 4.
- Open the Allocation smart button on the receipt and reserve the incoming quantity for the created delivery.
- Validate the receipt. Observation: The system tries to print the reception report, but validation fails with `External ID not found in the system: stock.stock_reception_report_action`.

Cause of the issue:
=========================
In this [PR](https://github.com/odoo/odoo/pull/264299/changes#diff-79ba983a362d3fed28458daad5f2902e914fd731308157aa73f350f1597265a6L30-L35), the reception report action was removed, and the reception report is now printed through the Picking Operations report using `do_print_picking()`. However, the auto-print methods in both Stock and MRP still referenced the old reception report action, causing the `External ID not found` error when the reception report auto-print is enabled

After This Commit:
=========================
- The obsolete reception report reference is replaced with the appropriate reports in Stock and MRP, preventing the traceback and allowing users to complete validation when the reception report is enabled.
- The unnecessary context override in `_get_autoprint_done_report_actions()` was removed. It was replacing the context returned by report_action() with the `unnecessary default_production_ids`, which removed the `active_ids` generated by  report_action(). These IDs are required by the report renderer to identify the production records, and removing them caused a blank PDF report.

Forward-Port-Of: odoo/odoo#283194

**Steps to reproduce:**
  * Set up a French company and configure Peppol E-invoicing.
  * Install `account_edi_ubl_cii` module.
  * Create a company partner (customer) and set a **Reference** value on the company contact under 
     **Customer** -> **Settings** -> **Sales and Purchase**.
  * Create a child contact under that company and set a different Reference value.
  * Create an invoice using the child contact as the invoice partner and confirm the invoice.
  * Send it via Peppol.

**Observed Behaviour:**
  * The BuyerReference in the generated XML contains the reference of the parent
  (commercial partner) Instead of the child contact used on the invoice.

**Cause:**
  * The buyer reference was taken from the commercial partner instead of the
  invoice partner.

**Fix:**
  * Update the condition to use the invoice partner's reference when available;
  Otherwise, fall back on the commercial partner's reference.

opw - 6330649

Forward-Port-Of: odoo/odoo#283324
Forward-Port-Of: odoo/odoo#273700

**Steps to reproduce:**
 - Install l10n_din5008 and accountant.
 - Enable Snailmail from Accounting → Settings.
 - Create a German customer (Fiscal Country: Germany).
 - Create and post a customer invoice using the DIN5008 report layout.
 - Select Send by Post.
 - Enable Developer Mode and navigate to `Settings → Technical → Email → Snailmail Letters`.
- Open the generated letter and send it.

**Current behavior:**
The letter fails to be sent to Pingen with the following error: An error occurred when sending the document by post.
Error: ` The attachment of the letter could not be sent. Please check its content and contact the support if the problem persists.`

**Cause:**
For Snailmail documents, Pingen validates that the recipient address is located within the DIN5008 address window. The current l10n_din5008 report renders additional document information instead of the address in the address area, preventing the compliance validation to fail.

**Fix:**
When rendering the report for Snailmail, ensure that only the recipient address is displayed in the DIN5008 address window while suppressing the additional information that would otherwise occupy this area. This preserves the standard DIN5008 layout for regular reports while generating a Snailmail-compliant PDF that passes Pingen’s validation.

**Reference:** [Pignen Recipient Address Validation Rule](https://help.pingen.com/en/fix-and-enhance-letters/issue-with-recipient-address#040201)

Ticket [link](https://www.odoo.com/odoo/project.task/6387869)
opw-6387869

Forward-Port-Of: odoo/odoo#280320

Steps to produce:
---
- Install `website_sale_collect` module.
- Create and publish a product.
- Add it to the cart and proceed to checkout.
- Fill in the address and confirm.
- Select a `pick-up in store` delivery method.
- Click the edit icon on the contact details.
- Confirm without making any changes.

Issue:
---
- The pick-up point address appears as a selectable delivery address in
  the contact details list, which it should not.

Root cause:
---
- When a pick-up point is selected, `set_pickup_location` calls
  `_address_from_json` ([1]), which creates a child `res.partner` record with
  `type='delivery'` and sets `pickup_delivery_method_id` to identify it as a
  pick-up point address.

  Later, when the user returns to the address page, `_prepare_address_data`
  calls `_get_delivery_address_domain` ([2]) from `portal`. This method returns
  all child partners with `type='delivery'` without distinguishing between
  user-created delivery addresses and the auto-generated pick-up point addresses

  As a result, the pick-up point address incorrectly appears in the checkout
  address list.

Solution:
---
- As specified in [task], partners created through this flow should be archived.
  However, in the referenced [commit], the `active=False` flag was removed
  when creating the partner, causing newly created partners to remain active.

  Override `_get_delivery_address_domain` to exclude pick-up point addresses.
  Since auto-generated pick-up point addresses always have
  `pickup_delivery_method_id` set, they are filtered out from the checkout
  address list, while manually created delivery addresses remain unaffected.

[1]https://github.com/odoo/odoo/blob/fb6298e50a7c8ded2800254e8715336eeb37deb5/addons/website_sale_stock/models/res_partner.py#L16-L72
[2]https://github.com/odoo/odoo/blob/fb6298e50a7c8ded2800254e8715336eeb37deb5/addons/portal/models/res_partner.py#L51-L55

[task]: https://www.odoo.com/odoo/project/49/tasks/3645144
[commit]: https://github.com/odoo/odoo/commit/fb74a371407ee19c6b1a3ab9f5a7b314978cb5cb

opw-6356778
---
I confirm I have signed the CLA and read the PR guidelines at www.odoo.com/submit-pr

Forward-Port-Of: odoo/odoo#283314
Forward-Port-Of: odoo/odoo#276148

Steps to reproduce
----------------------------------------
1. Install the SMS Marketing module (mass_mailing_sms).
2. Open any SMS Marketing record in the "Sent" or "Sending" stage.
3. Click on the "Insert Field" button.

Observation
----------------------------------------
Traceback Occurs:
```
TypeError: Cannot read properties of null (reading 'getRootNode')
```

Issue
----------------------------------------
The SMS widget displays the "Insert Field" button even when the SMS message
field is readonly. The button relies on the textarea reference to open the
dynamic fields popover, but the textarea is only rendered in editable mode.

The readonly behavior of the text field can be seen here:
https://github.com/odoo/odoo/blob/ccce9fcc79edcfb1f310b49a16de8235d987b74b/addons/web/static/src/views/fields/text/text_field.xml#L5-L7

However, the SMS widget still renders the "Insert Field" button without
checking whether the message field is readonly:
https://github.com/odoo/odoo/blob/ccce9fcc79edcfb1f310b49a16de8235d987b/addons/sms/static/src/components/sms_widget/fields_sms_widget.xml#L6

As a result, clicking the button in readonly mode tries to access an
unavailable textarea reference to open the dynamic fields popover, causing
a traceback.

Solution
----------------------------------------
Hide the "Insert Field" button when the SMS message field is readonly,
preventing the dynamic fields popover from being opened when the textarea
reference is unavailable.

---
I confirm I have signed the CLA and read the PR guidelines at www.odoo.com/submit-pr

Forward-Port-Of: odoo/odoo#283446
Forward-Port-Of: odoo/odoo#282845

Steps to Reproduce:

1. Confirm a Sales Order with one line -> creates delivery P1. Validate P1 with date_done = Day_A.
2. Create an invoice from the SO and post it -> invoice.delivery_date = Day_A.
3. Add a new line to the same SO -> creates delivery P2.
4. Validate P2 with date_done = Day_B, where Day_B is earlier than Day_A.

Issue:
The already-posted invoice's `delivery_date` silently changes from Day_A to Day_B after step 4, even though nobody edited the invoice. This only happens when a delivery validated after posting has an earlier `date_done` than what was already used.

Root Cause:
`account.move.delivery_date (sale_stock)` is computed in `_compute_delivery_date()`, which depends on `sale.order.effective_date.effective_date` is itself computed as the earliest `date_done` among all done, facing deliveries on the order. Neither compute method checks whether the invoice is posted, so validating P2 triggers a chain reaction: the delivery is saved -> the sale order recalculates -> the invoice recalculates -> delivery_date gets overwritten on an already-posted invoice.

`sale_stock` also marks `delivery_date` as protected, but this protection only works when the invoice itself is saved (write/create). Here, the change starts from saving the delivery (stock.picking), which never goes through the invoice's save method, so the protection never kicks in. `delivery_date` is also not on the list of fields Odoo normally blocks from editing after posting.

Fix:
`_compute_delivery_date()` now splits invoices into posted and non-posted before running. Non-posted invoices work exactly as before. Posted invoices are skipped from the sync and simply keep their current value instead of taking the newly calculated one. `sale.order.effective_date` itself is untouched only its effect on an already-posted invoice is blocked.

Result:
Once an invoice is posted, its `delivery_date` now stays fixed no matter what happens with later deliveries on the same sale order. `effective_date` keeps updating normally either way, confirming the fix only affects the invoice. Verified with both a script and a manual UI test.

opw-6409171

Forward-Port-Of: odoo/odoo#283069
Forward-Port-Of: odoo/odoo#280978

Current behaviour:
In the Calendar view (day/week/month scale), when the user's timezone observes a DST transition that starts exactly at local midnight (e.g. Africa/Cairo, since 2023), the day column right after the transition gets the wrong weekday name, duplicating the previous day's name. For ex. it renders "... THU THU FRI ..." instead of "... THU FRI SAT ...", for the week surrounding April 30th 2027.

To fix this we add 1 hour to the Date before reading its weekday/day from it, mirroring the workaround FullCalendar itself adopted for this same bug. It has no effect on any ordinary day (adding 1h to a correct local midnight stays within the same calendar day), and it cannot overshoot into the next day since no real-world DST gap exceeds that margin.

Note: This is a known bug (https://github.com/fullcalendar/fullcalendar/issues/7633), fixed in FullCalendar v6.1.17, a major version ahead of the v4.4.0, so the fix can't be applied directly without a full library upgrade.

opw-6370140

Forward-Port-Of: odoo/odoo#280253
Forward-Port-Of: odoo/odoo#279343

**Description**
The website analytics page always appears as if the website is not
connected to Plausible Analytics, even when the connection is actually
in place. This means that the user can't read analytics directly from
Odoo, and that the free Plausible Analytics offered is Odoo SaaS is not
usable.

**How to reproduce**
Set up Plausible Analytics (it is active by default on SaaS) and visit
`/odoo/website-analytics`, or go to "Reporting" -> "Analytics".

**Origin**
1. `WebsiteDashboard` component fetches analytics data via an RPC call to
   `fetch_dashboard_data`. For how the component is built, the first call
   misses a `website_id` argument.
2. `fetch_dashboard_data` defaults to `self.env.website` when called
   without `website_id` argument. After [1], `self.env.website` is empty
   on `jsonrpc` routes, because this protocol should be URL-agnostic.
As result, the `WebsiteDashboard` appears empty when first loaded, and
data is restored when a website button is clicked on the top right.

**Fix**
1. The argument `website_id` in `fetch_dashboard_data` is made optional.
2. `base.default_website` is used instead of `self.env.website`.

`base.default_website` is guaranteed to exist and to point to the first
website in the website sequence.

[1]: https://github.com/odoo/odoo/commit/5eba3e99ebd4f928fc5de620a38ab127e95be890

task-6402001

Forward-Port-Of: odoo/odoo#282003

### Issue before this commit:
When generating a SEPA Credit Transfer batch using the German XML format (pain.001.001.03), the <LEI> tag is erroneously included in the exported file if an LEI is configured on the company. This invalidates the XML, causing banks to reject the file.

### Steps to reproduce the issue:
1. Download Accounting and l10n_de
2. Go to Settins > Vendor Payments > SEPA Credit Transfer / ISO20022 and set Name Identification as 529900T8BM49AURSDO55 and Issuer as LEIMAN
3. Go to companies and set 529900T8BM49AURSDO55 as LEI in the DE company
4. Go to Accounting dashboard and click the 3 dots of the bank group, go to Configuration and set the Account Number and be sure in the Outgoing Payments tab XML Format is German
5. Create a new German company from Contacts with:
	1. Country as Germany
	2. VAT
	3. Account Number in the Bank Accounts by adding one line:
		1. example Account Number: DE65100500007201811026
		2. example Bank: BNP Paribas
		3. activate the Send Money button
7. Then go to Vendor > Payments and create a new one with Payment Method as SEPA Credit Transfer for the German company created
8. Go back and select the new payment from the list and click create batch and print it
9. In the XML of pain.001.001.03.(DE) file, the LEI tag should not be included.

### Cause of the issue:
The XML generation logic does not filter out the <LEI> element for older schema versions like pain.001.001.03, which do not support this tag.

### Reason to introduce the fix:
To ensure strict schema compliance and prevent bank rejections. The <LEI> element is now properly omitted from pain.001.001.03 files and restricted only to newer formats (e.g., pain.001.001.09) where it is valid.

opw-6428150

Forward-Port-Of: odoo/enterprise#128709
Forward-Port-Of: odoo/enterprise#127758

Currently an exception is generated when the user tries to export (downlaod) a
`CSV` file of the `General Ledger` report as the below step:
- Install the `accountant` module with demo data
- Enable and change to the `French` language
- Go to `Comptabilité` > `Analyse` > `General Ledger`
- Click the cog icon in the menu > Click on `CSV`
- An error occurs in the log, and nothing is downloaded

Error: `TypeError: 'NoneType' object is not subscriptable`

This issue occurs after the recent refactoring changes in [1]. When
`_generate_csv_lazy_export` is called, it uses the `_()` method for
translation, which accesses `self.env`. However, at this point, the cursor
is already closed. As a result, when the code at [2] is reached from
`ormcache`, it raises the above error because
`model.env.transaction.ormcaches__` is `None` (code ref [3]).

This commit fixes the above issue by performing the translation
using the existing `handler` variable, which contains an environment
with the new cursor (see code ref [4]).


[1]: https://github.com/odoo/odoo/commit/13c3adf3a8b5ba6325190d6b9aea45fb8a6a8b2f
[2]: https://github.com/odoo/odoo/blob/5ef7829895b2e05650da394c1e35dfdc3a23c066/odoo/orm/cache.py#L111
[3]: https://github.com/odoo/odoo/blob/5ef7829895b2e05650da394c1e35dfdc3a23c066/odoo/orm/environments.py#L1012
[4]: https://github.com/odoo/enterprise/blob/912a47b8f0828ef8316b7e4ecdabf8a2f305b313/account_reports/models/account_general_ledger.py#L524

Sentry-7608119520

Forward-Port-Of: odoo/enterprise#126778

**Problem:**
The running timesheet timer in the systray stops when the Odoo server clock passes UTC midnight, even though it is still the same day in the user's own timezone.

**Steps to reproduce:**
1. Set the user's timezone to one behind UTC (e.g. America/Guadeloupe, UTC-4).
2. Start a timesheet timer while it is before local midnight but after the server has passed UTC midnight (e.g. 20:00 local = 00:00 UTC).
3. Look at the running timer in the systray.

**Current behavior:**
At server (UTC) midnight the running timer disappears and its ongoing count is lost.

**Expected behavior:**
The timer keeps running until the user's own local midnight, regardless of the server timezone.

**Cause of the issue:**
The systray controller derives its reference day from `date.today()`, which returns the server's (UTC) local date. The running timer's timesheet is created dated the user's local day (`hr_timesheet` uses `fields.Date.context_today`). Once the server crosses UTC midnight, `date.today()` advances to the next day while the user's local day has not, so `timesheet_systray_user_data` (searching `date == today`) and `get_timer_start_time` (searching `date` within today's bounds) no longer match the running timesheet, and the systray reports no running timer.

**Fix:**
Deriving the reference day from `fields.Date.context_today` aligns the systray's notion of "today" with the timezone the timesheet was recorded in, so the timer is tied to the user's local day rather than the server's. This keeps recording and retrieval consistent, since the timesheet is already dated with the user-local day on creation.

opw-6343442

Forward-Port-Of: odoo/enterprise#125635

Currently, an error occurs when resetting a **bank** statement line to draft.

**Steps to Reproduce:**
 - Install the `Accounting` module.
 - Go to `Accounting Dashboard` and click the `three-dot` on bank journal.
 - Open `Transactions`.
 - Create a new `statement line`.
 - Select the `statement line`, click the `gear action`, and click `Reset to Draft`.

`AttributeError: 'bool' object has no attribute 'setdefault'`

After the [recent commit], when resetting the statement line to draft, the server action run [1] and 
the linked move is going reset to draft, and the method returns the result [2]. After the mentioned 
commit, the method returns True [3]. When the result from [4] is passed to clean_action, it raises an error [5].

This commit ensures that it returns None after resetting the statement line linked to the 
invoice to draft, as it previously returned None and same as like [6].

[recent commit]: https://github.com/odoo/odoo/commit/712718d9df0fd5044ac57fdd9ec58e64bece36c0

[1]- https://github.com/odoo/enterprise/blob/7ca28a1c079d22b60e3756ca9b4f404771f214e8/account_accountant/views/bank_rec_widget_views.xml#L541-L551

[2]- https://github.com/odoo/enterprise/blob/eae51e3ca155ca29e1de54f4bd223e7540b2aa7f/account_accountant/models/account_bank_statement.py#L112-L114

[3]: https://github.com/odoo/odoo/blob/b3444e4bd421a30229b0dc7c8d5eb6c78b8b51ed/addons/account/models/account_move.py#L6236-L6251

[4]: https://github.com/odoo/odoo/blob/a6f99706c6a62fc65666a0ff5e58fa465b41a6fb/addons/web/controllers/action.py#L53-L59

[5]: https://github.com/odoo/odoo/blob/b3444e4bd421a30229b0dc7c8d5eb6c78b8b51ed/addons/web/controllers/utils.py#L24

[6]: https://github.com/odoo/odoo/blob/d7df2e8acff9eb7066993fa6a0b0c6d7c85baabc/addons/account/models/account_payment.py#L1206-L1208

sentry-7354160052

Forward-Port-Of: odoo/enterprise#127189

**Steps to reproduce:**
* Install the **Accounting** module.
* Create two companies with different currencies (e.g. **USD** and **CAD**) sharing the same **Chart of Accounts**.
* Open a shared account and:
  * Add both companies in the **Company** field.
  * Under the **Mappings** tab, configure a mapping for each company.
* In each company, create and post a journal entry on the same shared account (for example, a receivable account) using the company's own currency.
* Set the journal entry dates to the **current month**.
* Open **Accounting → Reporting → General Ledger**.
* Change the reporting period to the **following month** so the posted entries are shown as the **Initial Balance**.
* Open the report separately for each company.

**Observed behavior:**
* From the **CAD company**, the Initial Balance displays **USD 2,000** instead of the expected **USD 1,000**.
* From the **USD company**, the **Currency** column on the Initial Balance is **blank**.

**Cause:**
* The SQL query for the `id_with_accumulated_balance` groupby used `SUM(amount_currency)` and `MIN(currency_id)` to aggregate all pre-period lines into a single Initial Balance row.
* In a multi-company shared Chart of Accounts, lines from different companies (each with their own currency) were collapsed into the same group, causing `SUM(amount_currency)` to add amounts across currencies and `MIN(currency_id)` to return an arbitrary currency ID.
* Additionally, the Python accumulation loop incorrectly performed **integer addition** on `currency_id` (a foreign key), further corrupting the displayed currency.

**Fix:**
* Replace `SUM(amount_currency)` and `MIN(currency_id)` with `CASE` expressions `MIN = MAX` is a uniformity check that works for **any number of currencies**: if every row in the group shares the same currency the condition is true and the correct sum is returned; if even one row differs the condition is false and both fields return `NULL`. The original three-column `GROUP BY (id, date, account_id)` is preserved.
* The Initial Balance row now correctly shows a **blank** currency column, consistent with the Odoo 18 behavior, instead of an incorrect aggregated foreign currency amount.

opw-6375310

Forward-Port-Of: odoo/enterprise#124823

Currently, an error occurs when orders are being fetched from shops with pending authorization.

Steps to replicate:
- Install `sale_tiktok`.
- Open Sales > Configuration > Shops (Under the title tiktok shops).
- Click `Connect New Shop` > Give values for `App key, App secret, Service ID`.
- Click `Connect Shop & Authorize` and then Return back to Odoo.
- Run the Scheduled Action `TikTok Shop: sync orders`.

Error:
```
File '/home/odoo/src/enterprise/saas-19.4/sale_tiktok/utils.py', line 171, in make_tiktok_api_request
    if now > shop.access_token_expire_datetime - timedelta(minutes=5):
TypeError: unsupported operand type(s) for -: 'bool' and 'datetime.timedelta'

ValueError: TypeError('unsupported operand type(s) for -: 'bool' and 'datetime.timedelta'') while evaluating
'model._sync_orders()'
```

Cause:
- Since the shop has not yet been authorized with TikTok, the `access_token_expire_datetime` field is not set. This field is only populated after the shop is successfully authorized (see [this]).

- Later, when the `TikTok Shop: sync orders` cron runs, the flow reaches [here], where we checks whether the access token is expired and needs to be refreshed. At this point, `access_token_expire_datetime` is still False because the shop has not been authorized yet.

Solution:
- The orders should only be fetched from those shops that are authorized with TikTok.
- Used the `access_token` field to determine whether a shop is authorized, as it is only populated after the authorization flow is successfully completed.

[this]: https://github.com/odoo/enterprise/blob/593409ca160863d3e985f3cf5e2aadda1d450b41/sale_tiktok/controllers/onboarding.py#L52-L54

[here]: https://github.com/odoo/enterprise/blob/593409ca160863d3e985f3cf5e2aadda1d450b41/sale_tiktok/utils.py#L171

sentry-7631179329

Forward-Port-Of: odoo/enterprise#127148

Creating a batch of payslips mixing employees with and without leave allocations raises a KeyError:
```
  File "l10n_au_hr_payroll/models/hr_payslip.py", line 869, in _add_unused_leaves_to_payslip
    annual_gross = leaves_totals[payslip.id]['annual'] * daily_wage
                   ~~~~~~~~~~~~~^^^^^^^^^^^^
  KeyError: 22
```

Current Issue:
`_l10n_au_get_unused_leave_by_type` only materialises leaves_by_date[payslip.id] inside the allocation loop, so a payslip whose employee has no matching allocation never gets a key. `_l10n_au_get_unused_leave_totals` then rebuilt a plain dict out of those entries and only fell back to a defaultdict when leaves_by_date was completely empty. A mixed batch is not empty, so the plain dict was returned and `_add_unused_leaves_to_payslip` raised on the payslips that were missing from it.

This never showed up in the UI, **where payslips are created one at a time**: a single slip either has an allocation, or produces an empty mapping that hits the fallback.

Approach:
Build the totals on a defaultdict and update it instead of returning a plain dict, so any payslip without allocation resolves to 0 rather than being absent. This also drops the need for the empty special case, and keeps the mapping consistent with the defaultdict returned by `_l10n_au_get_unused_leave_by_type`, which `_l10n_au_get_leaves_for_withhold` indexes the same way.

task-6465229

Forward-Port-Of: odoo/enterprise#127623

When reconciling an internal transfer between a branch and its parent
company, an User Error is raised if the branch transaction has been
processed via reconciliation model.

Steps to reproduce:
- Have a company with branch both selected
- On the branch, create a reconciliation model "Internal transfer" that
  assigns the whole balance to the liquidity transfer account
- Have a Bank journal on the company and a Bank journal on the branch
- On the branch bank journal, create a -100 transaction 'testb' and
  reconcile it using the branch internal transfer model
- On the company bank journal, creata a 100 transaction, open the
  reconciliation widget and select the branch transaction to match it

Issue:
The reconciliation is refused with a company inconsistency error
``` Uh-oh! You’ve got some company inconsistencies here:
- “BNK1/2026/00011 test” belongs to company “YourCompany” while
  “Reconciliation Model” (reconcile_model_id: 'Internal Transfer
  branch') belongs to another company.  To avoid a mess, no company
  crossover is allowed!
```
However, if user manually assign the transfer account to the branch transaction,
the reconciliation proceed as expected

Analysis:

When reconciling, we build the counterpart journal item by cloning the
values of the matched move line, copying also the reconcile model.
That field is company dependent and flagged copy=False, so it should not
be propagated.

opw-6365856

Forward-Port-Of: odoo/enterprise#127517

### Issue:
When an invoice has an image as its main attachment added via a log note, any non-admin user who did not create the attachment gets an `AccessError` when exporting the Datev ATCH zip

### Cause:
Since commit `e7c93e5a6f`, attachments uploaded via certain flows can be "orphaned" — their `res_model` is set to `False` and `res_id` to `0` via `_fix_attachments_on_record_from_files_data`
This allows the attachment to appear in the chatter without being linked to the move's attachment list

However, `_message_set_main_attachment_id` can still set such an orphaned attachment as `message_main_attachment_id`
When a user without system rights tries to read it, the ORM access check uses `res_model=False` and `res_id=0`, which does not match the move the user has access to, raising an `AccessError`

### Steps to reproduce:
- Install `l10n_de_reports` and switch to the DE company
- Create and confirm an Invoice (any lines, any customer)
- Add a Log Note with an image
- Set Demo user's Accounting rights to `Invoicing & Banks`
- Log in as Demo
- Open the General Ledger
- In the cog menu, choose `Datev ATCH (zip)`

Before the fix, an `AccessError` is raised

opw-6397804

Forward-Port-Of: odoo/enterprise#128484
Forward-Port-Of: odoo/enterprise#127787