Tuesday, September 8, 2026
4 changes · saas-18.3
Enhancements to existing features
Opening a Point of Sale register for a newly created company is now much faster when the database contains very large accounting histories from other companies. The change avoids slow database searches, reducing delays from seconds to milliseconds in large multi-company environments.
Original PR description
When you create a new company in a database that has a lot of existing account move lines and you attempt to open a PoS register from the list view, `_compute_company_has_template` checks…
When you create a new company in a database that has a lot of existing
account move lines and you attempt to open a PoS register from the list
view, `_compute_company_has_template` checks `_existing_accounting` for
the new company and wil run a sequential scan on the entire account_move_line
table followed by a nested loop as the query planner assumes AMLs company_ids
will be roughly evenly distributed.
This is not the case in a new company that has no/very few AMLs.
This is because the query ran is:
`SELECT COUNT(*) FROM
(SELECT FROM "account_move_line"
WHERE (
"account_move_line"."company_id" IN
(SELECT "res_company"."id" FROM
"res_company" WHERE
("res_company"."parent_path" LIKE '3/%')
))
LIMIT 1)`
and the values of company_id being searched for aren't known until the
subquery runs.
Running a query more like
`SELECT COUNT(*) FROM
account_move_line
WHERE company_id IN (%s)`
is much faster
Since res_company will always be a smaller table, we can do the inexpensive
search first and then pass in the values so Postgres can do a cheaper
search and return faster.
Benchmark time of `_existing_accounting`:
| Company 1 AML count | Company 2 AML count | Pre-fix | Post-fix | Multiplier |
|---|---|---|---|---|
| 10,000,000 | 0 | 700 Milliseconds | 500 Microseconds | 1,400x |
| 20,000,000 | 0 | 1.35 Seconds | 1 Millisecond | 1,350x |
| 20,000,000 | 20,000,000 | 2 Milliseconds |1.3 Milliseconds | 1.5x |
| 50,000,000 | 0 | 3.25 Seconds | 1 Millisecond | 3,250x |
| 100,000,000 | 0 | 5.3 Seconds | 1.3 Milliseconds | 4,075x |
Query Plan Before:
```
"Aggregate (cost=0.08..0.09 rows=1 width=8) (actual time=5573.001..5573.002 rows=1.00 loops=1)"
" Buffers: shared read=571435"
" -> Limit (cost=0.00..0.08 rows=1 width=0) (actual time=5572.995..5572.997 rows=0.00 loops=1)"
" Buffers: shared read=571435"
" -> Nested Loop (cost=0.00..https://github.com/odoo/odoo/commit/1571439c0c70c1f1dc3229421e696b97ce1678f8.33 rows=20000086 width=0) (actual time=5572.988..5572.989 rows=0.00 loops=1)"
" Join Filter: (account_move_line.company_id = res_company.id)"
" Buffers: shared read=571435"
" -> Seq Scan on account_move_line (cost=0.00..971432.72 rows=40000172 width=4) (actual time=0.432..1913.934 rows=40000000.00 loops=1)"
" Buffers: shared read=571431"
" -> Materialize (cost=0.00..4.03 rows=1 width=4) (actual time=0.000..0.000 rows=0.00 loops=40000000)"
" Storage: Memory Maximum Storage: 17kB"
" Buffers: shared read=4"
" -> Seq Scan on res_company (cost=0.00..4.03 rows=1 width=4) (actual time=0.785..0.785 rows=0.00 loops=1)"
" Filter: ((parent_path)::text ~~ '3/%'::text)"
" Rows Removed by Filter: 2"
" Buffers: shared read=4"
"Planning:"
" Buffers: shared hit=574 read=77"
"Planning Time: 15.323 ms"
"Execution Time: 5573.060 ms"
```
Query Plan After:
```
"Aggregate (cost=4.46..4.47 rows=1 width=8) (actual time=1.972..1.973 rows=1.00 loops=1)"
" Buffers: shared read=3"
" -> Limit (cost=0.44..4.46 rows=1 width=0) (actual time=1.967..1.968 rows=0.00 loops=1)"
" Buffers: shared read=3"
" -> Index Only Scan using account_move_line__company_id_index on account_move_line (cost=0.44..4.46 rows=1 width=0) (actual time=1.965..1.966 rows=0.00 loops=1)"
" Index Cond: (company_id = 3)"
" Heap Fetches: 0"
" Index Searches: 1"
" Buffers: shared read=3"
"Planning:"
" Buffers: shared hit=3"
"Planning Time: 0.219 ms"
"Execution Time: 1.998 ms"
```
opw-6513885
Forward-Port-Of: odoo/odoo#285096Resolved issues and error corrections
This fixes an issue where loyalty or coupon discounts in the Point of Sale could incorrectly reduce customer tips added during restaurant payments. Tips are now excluded from discount calculations, helping ensure staff gratuities remain accurate and unaffected by promotions.
Original PR description
Currently, when a discount reward is applied onto an order and a tip is added later, the discount gets also applied on the tip. Steps to reproduce: ------------------- * Make sure all demo data…
Currently, when a discount reward is applied onto an order and a tip is added later, the discount gets also applied on the tip. Steps to reproduce: ------------------- * Make sure all demo data loyalty programs can also apply to the restaurant * Make sure tipping is enabled in the restaurant * Open restaurant * Open a table, add items to the order * Activate coupon code "10pc" * Select payment * On payment page, add a tip (10$) * Go back to product page > Observation: 10% discount is also applied on the tip Why the fix: ------------ A tipping should never have a discount applied on it. During the computation of the discountable amount we have 3 distinct flow (discount applies on the order, on the cheapest line, on some specific products) - On the order: We always exclude tipping lines - Cheapest line: When fetching the cheapest line amongst all lines we don't consider the tipping lines - On specific: With the fix, tipping lines will always be excluded from the discountable lines. Even if the tip product was specified on the loyalty program. Should we still allow discounting tips if it was set in the program specifically? We're using `is_discountable` instead for `is_tip` in case we want to extend the definition later on. opw-6445364 Forward-Port-Of: odoo/odoo#281400
This fix checks for an existing proxy user before requesting new credentials from Odoo's IAP service. It prevents rare concurrent registrations from leaving a database with stale credentials that break later electronic invoicing proxy communication.
Original PR description
In the current flow, _register_proxy_user first calls IAP create_user, then inserts the returned credentials in account_edi_proxy_client.user. At the same time, IAP create_user_2 may replace an existing user with a new one with different credentials before local persistence settles. So Tx A calls IAP and gets credentials for remote user U1. Tx B calls IAP and gets credentials for remote user U2 and unlinks U1. Tx A persists U1 locally. Tx B fails local insert due to a unique constraint. Client DB keeps U1 credentials, but IAP now expects U2. Subsequent proxy calls from the client fail. The DB is left with stale, unusable credentials and cannot recover without re-registration. IAP: https://github.com/odoo/iap-apps/pull/1816 task-6520501 --- I confirm I have signed the CLA and read the PR guidelines at www.odoo.com/submit-pr Forward-Port-Of: odoo/odoo#285648
Fixed an issue where one task that could not be scheduled correctly caused other tasks assigned to the same person to be pushed far into the future. This keeps dependent project work scheduled in the next available slot, improving planning accuracy in Gantt rescheduling.
Original PR description
Steps to reproduce: 1. create three tasks A,B & C with the same assignee 2. make tasks B and C depend on A and start on the `date_deadline` of task A 3. set the duration of task B (processed before C) to be longer than the 53-week search window 4. reschedule task A to end after task C starts Problem: A candidate that couldn't be scheduled was putting its assignee in conflict, subtracting the inspected availability while failing from the shared pool. Consequently, every later candidate sharing that assignee got force-placed into the same forward reschedule fallback that lands near the tail of the 53-week search window instead of the next available slot. Solution: Each candidate should be evaluated on its own ability to fit, and the time it occupies should count against the shared pool. opw-6380163 --- Forward-Port-Of: odoo/enterprise#129326