Steps:
- Install ai
- Open ai chat
- Try to send a message
- Traceback

The AI chat was failing on iOS with a traceback when trying to process
the stream response.

Even though MDN suggests compatibility, Safari (WebKit) versions prior
to 26.4 do not implement the AsyncIterator protocol on ReadableStream.
This makes `for await (const chunk of response.body)` throw a TypeError
as `[Symbol.asyncIterator]` is undefined.

https://developer.mozilla.org/en-US/docs/Web/API/ReadableStream

This commit replaces the async iteration with a manual reader loop
(`getReader().read()`). This is the low-level primitive supported by
all versions of Safari and ensures the stream is properly consumed and
unlocked even if the connection is interrupted.

Steps to reproduce:
1. Open Discuss/AI Chat on an iPad or iPhone (eg. 26.4).
2. Send a message.
3. The response triggers a JS error.

```js
Uncaught Promise > undefined is not a function (near '...chunk of asyncStream...')
```

opw-6054307

Forward-Port-Of: odoo/enterprise#113926

### Issue before this commit:
The "Source" (origin) field was missing from the PDF of replacement invoices. While the original invoice correctly displayed the Sales Order reference, the new invoice generated through the request cancel process had an empty origin field.

### Steps to reproduce the issue:
1. Download Sales and l10n_mx
2. Set a UNSPSC Category for one product
3. Go to Sales, create a new Quotation and confirm it
4. Create invoice, confirm and send & print
5. Request cancel button -> create replacement invoice
6. In the new invoice there is no source origin invoice

### Cause of the issue:
The invoice_origin field is defined with copy=False. Since the replacement logic uses the copy_data method without explicitly passing the origin value, the field was automatically cleared during the creation of the new invoice.

### Reason to introduce the fix:
To ensure document traceability, the fix explicitly passes the invoice_origin from the original invoice to the replacement. This maintains the link to the Sales Order in the database and ensures the "Source" label appears on the printed PDF.

opw-6070016

Forward-Port-Of: odoo/enterprise#114099

Steps to reproduce Bug 1:
  1. Login as a user with "All Timesheets" (Approver) access.
  2. Disable the "Timesheet Assistant" feature for this user.
  3. Ensure "Billing Rate Indicators" is enabled in settings.

Steps to reproduce Bug 2:
  1. Login as a user with "All Timesheets" (Approver) access.
  2. Disable the "Billing Rate Indicators" setting in company settings.
  3. Ensure "Timesheet Assistant" is enabled in settings.
 
Steps to reproduce Bug 3:
 1. Only install 'sale_timesheet_enterprise'.
 2. Go to Timesheets > Configuration > Settings.
 3. Toggle "Billing Rate Indicators" (timesheet_show_rates) or change the encoding unit (timesheet_encode_uom_id), then save and check the menus.

Issue:
1. The "Configuration" menu is hidden, preventing access to billing targets even if the user has "All Timesheets" access.
2. The "Billing Time Targets" menu is still visible inside Configuration even if the "Billing Rate Indicators" feature is disabled in the settings.
3. Menu visibility does not update immediately after saving the settings. Menus that should appear (e.g., "Employee Billing Time Targets" or "Timesheets Assistant") remain hidden, or vice versa, until the cache is cleared or the server is restarted.

Cause:
1. The `hr_timesheet_enterprise_menu_configuration` was restricted in XML to groups that excluded "All Timesheets" users.
2. The `_load_menus_blacklist` logic in Python only blacklisted billing menus for users who were both Managers and System Admins, leaving them visible to regular Approvers even when the feature was disabled.
3. The load_menus method is decorated with @ormcache and stored in the Registry LRU cache. Menu visibility depends on timesheet_show_rates and timesheet_encode_uom_id through _load_menus_blacklist. When this field is updated, the ORM does not automatically invalidate the cached load_menus result because these specific fields are not part of the configuration fields. As a result, the stale old menu remains in memory.

Fix:
- Updated XML to include `hr_timesheet.group_hr_timesheet_approver` in the Enterprise Configuration menu permissions.
- Refactored `_load_menus_blacklist` to:
    - Hide all billing-related menus for all users when the feature is disabled.
    - Hide the parent Configuration menu if it would otherwise be empty.
- Override the write method in res.company in both modules and explicitly call env.registry.clear_cache() when the relevant configuration fields are modified.

task-5428010

Forward-Port-Of: odoo/enterprise#106464

### Issue:
If a payslip is split into multiple bank accounts (Salary Allocation), the generated SEPA file contains duplicate <InstrId> tags

### Cause:
The `_get_payments_vals` method, `InstrId` is based on the payslip ID
When a single payslip generates multiple transaction blocks, this ID is duplicated, violating the ISO 20022 requirement for unique instruction identifiers
https://knowledge.xmldation.com/support/iso20022/general_rules/instrid

This commit adds a unique suffix (e.g., -1, -2) to the `InstrId` for each transaction generated from the same payslip to ensure technical uniqueness Nothing change when you only have one account

This is the part of the code that use the payment name:
https://github.com/odoo/enterprise/blob/194a8d35ef3e9b47ff566479b0c35c0f963fb42d/account_iso20022/models/account_journal.py#L294-L299

### Steps to reproduce:
- Install `hr_payroll_account_iso20022` with demo data
- On the Bank Journal, set a valid IBAN (e.g. BE04957751619131) for `Bank Account Number`
- Open the Employee page for Abigail Peterson
- In the Personal tab, add 2 Bank Accounts (Send Money: True, Account Number: any)
- Click on Salary Allocation and Save (You'll have a 50/50 ratio)
- Create a new Pay Run (for Abigail Peterson)
- Open the last PaySlip and Validate
- Create Payment Report (Export Format: SEPA)
- Download the Payment Report and check the <InstrId> tags

opw-6069670

Forward-Port-Of: odoo/enterprise#114821
Forward-Port-Of: odoo/enterprise#113113

Problem: Portal email with auto-assignment crashes ticket creation due to calendar access.

When a helpdesk ticket is created via email from a portal user and automatic assignment is enabled, the system computes working intervals for users to determine assignment.

This computation goes into resource logic, where resource.calendar fields (flexible_hours) are read.

If the assigned user is linked to multiple employees across companies, multiple resource.resource records are evaluated.

The helpdesk email flow starts in sudo, but the employee calendar lookup explicitly drops sudo before returning the calendar. Then, the calendar is accessed in the portal context, which does not have permission to read the other company's resource.calendar, leading to an AccessError and preventing ticket creation.

Although the failure is triggered from Enterprise helpdesk, the actual crash occurs in Odoo (resource.calendar), meaning the fix must be applied there.

Fix: Preserve sudo when fetching employee calendars to ensure that scheduling logic does not depend on the access rights of the email sender.

A test is added in helpdesk_holidays, as the issue requires both helpdesk (auto-assignment) and hr (employees/resources) to reproduce. The test simulates a portal email flow with a multi-company user linked to multiple employees and ensures ticket creation succeeds.

Steps to Reproduce:
1. Install Helpdesk, Employees, and enable multi-company
2. Create two companies (e.g., Company A and Company B)
3. Create one internal user (User X) with access to both companies
4. Create two employees linked to the same user:
- Employee 1 in Company A
- Employee 2 in Company B Make sure they are set with a start date, but no end date. Needs to be active employee.
5. Create a Helpdesk team in Company A
6. Add Agent X as a team member
7. Enable automatic assignment
8. Configure an email alias for the helpdesk team
9. Create a portal user
10. Send an email from the portal user to the alias

Related Ticket: opw-6035099

Forward-Port-Of: odoo/enterprise#114986
Forward-Port-Of: odoo/enterprise#113047

When a company has a shortened fiscal year defined via account.fiscal.year (e.g. May-December), the depreciation board computation for degressive assets incorrectly computes the start of the next fiscal year using `date_from + 1 year` instead of querying the actual next fiscal year. This causes entries for the months between the wrong and correct FY start (e.g. January-April) to be skipped entirely.

Step to reproduce:
- Create a company with a fiscal year starting in May (e.g. May 1st 2025 to 31st December 2025)
- Create an asset with a start date in the 1 December 2025, with a 24 months duration and degressive method
- Compute the board and observe that entries from January to April 2026 are missing

Fix the FY boundary detection in _recompute_board to query the fiscal year containing the day after the current period end, revert the effective_start_date logic in _compute_board_amount that was masking the root cause, and move the prorata date clamping to _create_move_before_date where it is needed for disposal.

opw-6016834

Forward-Port-Of: odoo/enterprise#113895
Forward-Port-Of: odoo/enterprise#113521