Skip to content

Live comments and comment creation race against each other #2458

New issue
New issue
Open
#2462
Open
Live comments and comment creation race against each other#2458
#2462
Assignees
Soxasora
Labels
bug
@Soxasora

Description

@Soxasora
Soxasora
opened on Aug 28, 2025

Is there an existing issue for this?

  • I have searched the existing issues

Describe the bug

Comments are created and injected into cache as a result of a paid mutation.
If I understood everything correctly, if the payment method is optimistic, the comment is immediately injected into the cache. However, if the payment method is pessimistic, such as a QR payment, the comment is only injected into the cache after the payment has been confirmed.

Only paid new comments are fetched, but a pessimistic mutation delays cache updates, causing double comments

For anons, edits are based on the payment hash stored in local storage after the payment has been confirmed

stacker.news/components/use-paid-mutation.js

Lines 130 to 145 in c5b50d7

// the action is pessimistic
try {
// wait for the invoice to be paid
// returns the invoice that was paid since it might have been updated via retries
invoice = await waitForPayment(invoice, { alwaysShowQROnFailure: true, persistOnNavigate, waitFor, updateOnFallback })
if (!response.result) {
// if the mutation didn't return any data, ie pessimistic, we need to fetch it
const { data: { paidAction } } = await getPaidAction({ variables: { invoiceId: parseInt(invoice.id) } })
// create new data object
// ( hmac is only returned on invoice creation so we need to add it back to the data )
data = mergeData({ ...paidAction, invoice: { ...paidAction.invoice, hmac: invoice.hmac } })
// we need to run update functions on mutations now that we have the data
update?.(client.cache, { data })
}
ourOnCompleted?.(data)
onPaid?.(client.cache, { data })

stacker.news/components/use-item-submit.js

Lines 76 to 79 in c5b50d7

onCompleted: (data) => {
onSuccessfulSubmit?.(data, { resetForm })
paidMutationOptions?.onCompleted?.(data)
saveItemInvoiceHmac(data)

This implies that if we were to skip a cache update when creating a comment if it’s already in the cache, the canEdit conditions won’t be fulfilled because live comments will inject it before the pessimistic mutation is complete, necessitating a refresh.

Screenshots

anonlivecomments.mp4

Steps To Reproduce

  1. be anon
  2. create a comment right when live comments fetching occurs
  3. comment is duplicated and the fetched one can't be edited.

This requires timing precision or luck, can be reproduced more reliably with a 1 second polling rate in dev.

Expected behavior

Comment is created and live comments won't inject it.

Logs

No response

Device information

No response

Additional context

Reported by @ekzyis
This couldn't occur in the first version of live comments because of multiple steps of deduplication that delayed injection. The permanent fix shouldn't be more steps of deduplication in live comments, maybe we should deduplicate on comment creation if it's a pessimistic mutation

Metadata

Metadata

Assignees

Labels

bug

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions