With:

Two companies 'Company A' & 'Company B'
Sequences for SO & PO for each company
so_from_po set to True on 'Compny B'
When:

Company A created a PO with Company B as supplier, and validated it,
An error is raised "Configure correct warehouse for company(Company B)..."
The field warehouse_id is required on the SO, so force the user to set one in
the configuration of the company

account.bank.statement.line has a journal_entry_ids field which is a o2m
to account.move. During confirmation of bank
statements (button_confirm_bank()) this o2m field is heavily used. When
a database has a lot of account.move records this lookup starts taking
up quite a bit of time. As an example, on a locally duplicated customer
database with ~700000 account.move records attempting to resolve
journal_entry_ids of an account.bank.statement.line record was observed
to take > 1 second. This index improves it to be < 5 milliseconds.

This is especially important when big POS sessions have to be closed,
because those will contain a lot of account.bank.statement.line records.

opw-665309

When editing a form with the arabic language, the date is formatted in
arabic. But the version of the Moment.js library we use for this had
inconsistencies and parsing the arabic date could not work.

Hence, saving would not work since date field values would be invalid.

This commit cherry-pick http://momentjs.com/ commits which solves that:

  from version 1.8.4: https://github.com/moment/moment/commit/c9023b5
  from version 1.9.0: https://github.com/moment/moment/commit/bd328d5

closes #10634
opw-660453

When creating a XLS file with more than 256 columns, the library used
(xlwt) fail since it is targetted to support MS Excel 97 up to
Excel 2003.

But Odoo doesn't has no check and in this given instance (an error
happening in a controller generating a binary filte) the real error
message is lost.

This commit check if the to-be exported data has more than 256 columns,
and if this is the case display an error message without even trying
the export.

closes #10630

opw-660474

[FIX] search with computed fields in path

fields in a search path

When performing a stock valuation at date,
the stock valuation total wasn't equal
to the sum of each line of the report.

This is because the domain forcing the
date wasn't passed to the `search` call
when no group by was applied.

Indeed, when calling
`read_group` with a group by, the lines in the results
contains the domain used for each line, but when
there is no group by applied, this is not the case, the domain
is not included in the line dict returned by `read_group`.

In such a case, therefore, we must use the domain that was passed to
the `read_group` call.

opw-667761

Backport of fda5021e that I
erroneously did in 9.0, thinking it was reported on 9.0.

In a survey question of type `matrix`,
nothing prevents to remove a row from the matrix,
even if there are already answers for this line.

If the case happens, the row is deleted, but
not the answers. The answers are therefore orphan.
The answers should probably be deleted when the
row is removed from the survey, but this is a risky
change, and, even, users might want to keep track
of the answers given even if the line doesn't exist
anymore in the survey.

Therefore, we just handle the case when displaying
the survey results,
we ignore matrix answers for which the row no longer
exists.

opw-666393

Accessing a related field with `related_sudo=True` on a draft record should
effectively traverse the fields as the admin user.

This fixes #5121, #7138.

Suggested alternative descriptions spill out of the boxes if the
description is too long.

PS: same fix as in saas-6 with f7110b46.
This fix doesn't have to be forwrad ported.

opw:667343

When writing several times on the same record in the same transaction,
the orm raise a Missing error because the rowcount only gives one
difference.

opw:666470

If there is no contract then the default journal_id must be set.
PS: The field "journal_id" in "hr.contract" model is not required.

opw:667093

The `stock.picking`.`state` field is set
to track the change of values
(`track_visibility='onchange'`)

It's supposed to write the state changes
within the picking thread.

It does not work properly for function fields,
as the onchange tracking is designed to work
only with direct user changes, direct
`write` operations on the record, while,
here, the state value changes according to the
picking moves changes, for instance.

To solve this, the tracking changes have to be
hooked within the `create` & `write` methods
of the model on which this function depends on.

Therefore, from now, when changes are
performed in the moves, on the fields that
could lead to the picking state change, we
force the tracking of the picking state.

In addition, we had the `mail_notrack` key
in the context when creating back orders,
to avoid displaying the back and forths
in the state
e.g. when transferring 9 units on 10, the
changes were displayed as below:
Draft -> Waiting availability
Waiting Availability -> Ready to Transfer
Ready to Transfer -> Draft
Draft -> Partially available
Partially Available -> Waiting availability
Waiting Availability -> Transferred

opw-666317

When a PO is created automatically from a procurement, the currency is
the currency of the connected user, which might be different from the
currency of the pricelist.

The fix set the currency to the currency of the pricelist.

opw-665573

When creating a company, a new partner is created with his field
'customer' = True. But by default a company doesn't have to be
a customer.

opw:667130

When deleting the resource associated to a mail statistic,
e.g. deleting the lead associated to a statistic,
the statistic could not longer be opened.

opw-656738

When doing a copy of a google drive spreadsheet (in google_drive.py's
copy_doc method, using the following google API call
https://developers.google.com/drive/v2/reference/files/copy). Google
server return an error if we used the old spreadsheet KeyId.

The new FileId is available in the AlternateLink and this commit updates
it.

similar-to: 00c2a998

opw-647222

When a PO line is created from procurement, the supplier taxes set
on the PO line must be visible in the company set on the procurement.

opw:666875

Apple devices have a patented feature called rubberbanding that allows
the user to scroll past the end (or beginning) of a document. Browsers
on IOS have this same feature.

This is not an issue most of the time. But sometimes the rubberbanding
happens to the entire document.body of the POS, which is not
ideal. Rubberbanding is already disabled partially by
Chrome.disable_rubberbanding(). But it is not perfect. The most
important issue that remained is what happens when a scrollable element
is either at it's top or bottom and the user *initiates* a scroll past
its beginning or end. So scrolling an element all the way to the top and
then continuing to scroll did not trigger the problem. If however the
element was already at the top and you would initiate a scroll up the
scroll would fall through to the document instead and the entire POS
would scroll.

What aggrevates the issue from a UI point of view is that as long as the
rubberbanding animation is not finished, any subsequent scrolls will
target the entire document as well. Normal treatment of scrolling
resumes only when the animation has finished and the document is in the
center of the screen again.

Stopping this behavior is not trivial. The only know solutions are all
built around stopping touchmove events, which we already did.

This patch makes it so that we also stop all touchmove events that
trigger the unwanted behavior described above. It ensures that we stop
touchmove events that would scroll an element above it's top or below
it's bottom. To do this we have to manually determine whether or not the
element is scrolled all the way to the top or bottom and we have to
manually determine the direction the user is swiping.

closes #10488

This is related to revision 44eeb387

The origin concatenation was not working
when `procurement.origin` was `False`

opw-666933

The price displayed on the event page is the
`event.ticket`.`price_reduce` field,
which is basically the ticket price(`event.ticket`.`price`)
minus the possible discount applied by the pricelist

Nevertheless, the price asked when ordering the ticket,
in the cart/checkout, was the `ticket.price`,
without the possible discount from the pricelist, therefore.

The price asked for the ticket was therefore different than the price displayed.

To reproduce:
 1. Settings > Sales > Use pricelist
 2. Sales > Configuration > Pricelist > Public Pricelist > Apply a 20% discount (-0.2)
 3. Go to /event, -> Conference on Business Applications
 4. Order 1 of each
 5. Notice that the price asked is 800€ and 1200€ instead of 1000€ and 1500€ respectively
 6. Hit Order now -> Notice that the price in the cart are the price without the discount

opw-665540

Quants should never be changed or deleted by a user.
We make it not possible anymore in the views.
And also in the code we avoid to unlink without
passing a specific key in the context.
(as we still need to be able to delete quants in case of
negative quants reconciliation)

On iOS (at least version 8.3 and up), both Chrome (at least 47 and up)
and Safari emit ghost clicks [1] that added the product normally
rendered under the table widget to the order. So when selecting a table
without any orders you would end up with the table selected, and the
product under the table in the order after just tapping the table.

The issue is that on the table selection screen we listen to touchend
events. On the product selection screen we listen to click
events. Mobile browsers will usually emit both, first the touchend
event, then a small delay, and then the click event. The small delay is
platform dependent. Also, depending on a bunch of circumstances,
browsers adjust this delay time (see [1] for more info).

On Android this never caused issues because there the click is fired
pretty fast after the touchend and more importantly, the click event
still has the table widget as it's target. We don't listen to click on
the table widget so it gets correctly ignored.

On iOS there is a longer delay, and the click event is emitted with
whatever target is currently under the coordinates of the click.

In order to stop this from happening we'll just bind to click when not
editing. When we are editing the floor layout we'll do what we did
before.

We also bind to dragend because on touch devices it is sometimes not
easy to click, especially on small elements. You have to touch and
release the screen without moving your finger. Because the table
selection screen hides overflow there are no downsides to doing this,
and it makes selecting a table much easier and less likely to fail due
to accidental dragging.

Calling preventDefault on the touchend event is another option but did
not work on Android running Firefox 43.

[1] http://ariatemplates.com/blog/2014/05/ghost-clicks-in-mobile-browsers/

When displaying a dashboard, if a custom default filter was set on a
view it would be concatenated to a given dashboard filter.

Since it is never wanted and dashboard are embodied by fixed saved
filters, thus not needing custom filter, this fix simply disable them
when viewing the dashboards.

closes #10477

opw-660140

Field virtual_available on product.product is called Forecast Quantity but was
not on product.template.

This is not an issue to modify in stable as the term is already translated and
the translation of the string of the field is not based on the source (meaning
that, without reloading the translations, people will still see the translation
for 'Quantity Available').

Closes #10443