Cht-core: IDs of new documents are not added to changes requests until page refresh

[alxndrsn]

To replicate:

• Open a restricted user, watch the network. You should generate a longpoll _change request once everything is synced.
• Create a new document. Note that a bunch of stuff happens on the network, but specifically that the first longpoll is satisfied by the document you just created, and a new longpoll is created after this.
  
  
  
  • Do that a few more times for good effect.
  
  

What should happen:

• Every time a new longpoll request is initiated after you write a document, the id of that document should be among the ids passed

What actually happens:

• The ids are the same each time.
• If you refresh the longpoll that is initiated after this has the correct doc ids.

Looking at the code this is actually obvious: we only add the doc_ids when we're instantiating the two way sync the first time.

[SCdF]

I updated the description

[medic-bot]

Hi @abbyad,

Please ensure this ticket has a Priority, Status and Type label.

(See triaging old issues for more detail)

[abbyad]

Redirecting to @alxndrsn to put the type Status and Priority

[alxndrsn]

I'd guess this is low priority, as:

• deleting documents that have just been created by a different user probably isn't that common. But perhaps deletes aren't reaching all phones used in training?
• fixing it looks fiddly

Thoughts, @garethbowen @SCdF?

[SCdF]

Well I think ideally we'd fix this implicitly by redesigning the replication API so we don't have to know the ids at all, see: #4185

If we can't fix that particular oddity in the algorithm we should look into this.

[garethbowen]

We no longer include doc ids in the request so this is fixed.