PaymentRequest API is not supported when the PaymentRequest is not in the
fully active document. Adding bool a PaymentRequest::InFullyActiveDocument()
method to check if the PaymentRequest is in fully active document. This
method should be called at the start of any PaymentRequest APIs. If the
PaymentRequest is not in fully active document, ignoring the API call and
throw the NS_ERROR_DOM_ABORT_ERR if needed.
1. Add a new method nsIPaymentRequestService::ChangePaymentMethod for UI to
inform merchant the payment method is changed by user.
2. Add a new method in PPaymentRequest.ipdl to passing the changing method
information from chrome process to content process.
3. Add BasicCardChangeDetails dictionary in BasicCardPayment.webidl
4. Create ChangeDetails, GeneralDetails and BasicCardDetails for saving the
the method information in content process.
1. Add a new method nsIPaymentRequestService::ChangePaymentMethod for UI to
inform merchant the payment method is changed by user.
2. Add a new method in PPaymentRequest.ipdl to passing the changing method
information from chrome process to content process.
3. Add BasicCardChangeDetails dictionary in BasicCardPayment.webidl
4. Create ChangeDetails, GeneralDetails and BasicCardDetails for saving the
the method information in content process.
1. Adding a new completeStatus 'initial' for nsIPaymentRequest.completeStatus
to indicate the status of the showing PaymentRequest for Payment UI component.
2. Removing the PaymentRequest::mDeferredShow and calling
PaymentRequestManager::showPayment() when PaymentRequest::Show() called
with a detailsUpdate parameter to inform UI component to support better
user experience.
In original design, payment method response data is passed between processes
through a simple nsString. It means a special encoder/decoder is needed for
special response data, ex. BasicCardResponse, to serialize/deserialize
into/from the nsString. However, when a token spliter, ':', ';' and '@', is
used in response data, it makes the encoder/decoder can not work normally.
It is hard to define a suitable token spliter set for encoder/decoder.
So instead of using an error-prone encoder/decoder, this patch defining a new
IPC structure for response data.
1. Add PaymentValidationErrors and PayerErrorFields in PaymentRequest.webidl
and PaymentResponse.retry() in PaymentResponse.webidl
2. Implement PaymentRequest.retryPayment() and
PaymentRequestManager.retryPayment() in content process.
3. Add IPCPaymentRetryActionRequest in PPaymentRequest.ipdl to transfer the
error fields to chrome process.
4. Implement PaymentResponse.retry() by reusing the code of show() and
updateWith() of PaymentRequestService in chrome process.
1. Checking if the PaymentRequest::mState equals to eInteractive when calling
PaymentRequest::UpdatePayment() and PaymentRequest::AbortUpdate.
2. Removing the non-suitable MOZ_ASSERT in PaymentRequest::RespondShowPayment()
1. Returning NS_ERROR_FAILURE in PaymentRequestService::RespondPayment() when
the specified PaymentRequest does not exist in PaymentRequestService.
2. Renaming "cleanup" to "close" in DOM::WebPayment codes to fit the actual
behavior.
Let PaymentRequest inherit from nsIDocumentActivity interface.
Calling RegisterActivityObserver() and UnregisterActivityObserver() in
constructor and destructor to get activity notifications from document.
When receiving the notification, NotifyOwnerDocumentActivityChanged() will
check the owner document's activity status. If the status is disabled, calling
PaymentRequestManager::CleanupPayment() to cleanup the PaymentRequest in
content process and aslo sending the cleanup information to chrome process.
We only build the frontend parts of Web Payments on Nightly, but users could
futz with the prefs and expose the DOM API on Beta. We should be careful about
not allowing that mismatch.
This also adds some additional logging if we run into a bug like this again.
MozReview-Commit-ID: 9qcQTIsIHkg
If we end up shutting down the content process while there's a request active
we'll end up "leaking" the request and asserting. This makes sure that we
let go of the PaymentRequest in that edge case. I can cause this by forceably
closing the payment request window on Linux without hitting OK or cancel.
MozReview-Commit-ID: 6XDYIcqNkC6
The existing code only passed strings around to identify PaymentRequest
objects. That meant that we were constantly having to do hashtable lookups to
find the corresponding object. This patch just uses the single map that IPDL
maintains for us and passes objects around, obviating the need for more
hashtable lookups.
MozReview-Commit-ID: 6BBonrc6q4x
There isn't any need to create an actor per call to the parent. This patch
lines up PaymentRequest with PPaymentRequestChild objects and links them
together. It also simplifies the maps and arrays we use to keep track of these
objects.
There's one tricky bit to note in this patch: in the case that a promise is
passed to paymentRequest.show(), we don't notify the parent process until the
promise resolves (when we call either UpdatePayment or AbortUpdate). In that
case, I needed to distinguish between an "update" because of the promise
resolving or a call to updateWith on an shippingaddresschange event in order
to get the bookkeeping right with the mActivePayments hashtable. In that case,
the PaymentRequest is kept alive by mShowingRequest alone. In all other cases,
mActivePayments keeps the PaymentRequest alive until we resolve or reject the
correct promise.
MozReview-Commit-ID: HoHjn8eqC4T
This removes some hard tabs that crept in and combines two if branches into a
single statement in order to avoid a bit of duplication.
The existing code here seems to treat any sort of upcast as needing a QI. That
is needlessly wasteful and causes a bunch of unneeded virtual calls.
MozReview-Commit-ID: 7WshYm9C4Xb
1. Add "optional Promise<PaymentDetailsUpdate> detailsPromise" as a parameter
of PaymentRequest.show() in PaymentRequest.webidl.
2. Let PaymentRequest inherit from PromiseNativeHandler, and implement the
ResolvedCallback() and RejectedCallback() to handle the PaymentDetailsUpdate
promise.
3. Update PaymentRequest.show() implementation. If PaymentDetailsUpdate Promise
is not nullptr, the show request would not be transferred to chrome process
immediately until the promise is resolved/rejected.
4. Update selectedShippingOption when requestShipping is true.
5. Change the PaymentMethod id validation sequence according to the spec.
Add a new member mAllowPaymentRequest on nsIDocument and following related
methods,
bool AllowPaymentRequest() const
void SetAllowPaymentRequest(bool aAllowPaymentRequest)
mAllowPaymentRequest is used to check if PaymentRequest is allowed to use on
the document. According to the spec
https://html.spec.whatwg.org/multipage/iframe-embed-object.html#allowed-to-use
the mAllowPaymentRequest should be true under following situations
1. The document is the top level content document.
2. The document is the same origin with its parent document and its parent
document's mAllowPaymentRequest is true.
3. The document is different origin with its parent document but its
parent node is an iframe with allowpaymentrequest flag and the parent
document's mAllowPaymentRquest is true.
This patch also include following mochitests
1. test for allowpaymentrequest flag changing. The flag change effect should
not be applied to the document immediately, it should be updated while
the browsing context is updated.
2. test for effect propagation in the nested iframe.
1. PaymentRequest::PrefEnabled() returns false under non-e10s situation.
2. Add a new mochitest test_block_none10s.html to expect to get ReferenceError
in non-e10s mode.
