fix(event cache): keep the previous-batch token when we haven't enabled storage #4495
Conversation
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #4495 +/- ##
==========================================
- Coverage 85.36% 85.36% -0.01%
==========================================
Files 284 284
Lines 31886 31903 +17
==========================================
+ Hits 27219 27233 +14
- Misses 4667 4670 +3 ☔ View full report in Codecov by Sentry. |
This removes a few manual uses of `ResponseTemplate`, which is sweet and guarantees some better typing for those responses overall.
bnjbvr
force-pushed
the
bnjbvr/store-prev-batch-token-if-no-storage
branch
from
9023340 to
bb444d0
Compare
| MatrixMock { server: self.server, mock } | ||
| } | ||
| } | ||
|
|
||
| /// A response to a [`RoomMessagesEndpoint`] query. | ||
| pub struct RoomMessagesResponse { |
There was a problem hiding this comment.
Can we call this a bit differently? It's not the real Ruma response we use everywhere else, it's a test helper. Perhaps RoomMessageResponseTemplate?
There was a problem hiding this comment.
Well we're not mocking, and it's not a message. It's about the response that you would send out if you were to send a room message. Then we pass that response to the mock server which will mock things for us.
I picked the Template naming because wiremock, which we use and wrap here, is using that term as well: https://docs.rs/wiremock/latest/wiremock/struct.ResponseTemplate.html#.
There was a problem hiding this comment.
I like RoomMessageResponseTemplate, thanks!
| MatrixMock { server: self.server, mock } | ||
| } | ||
| } | ||
|
|
||
| /// A response to a [`RoomMessagesEndpoint`] query. | ||
| pub struct RoomMessagesResponse { |
When we don't have enabled storage, we should keep all the previous-batch tokens, otherwise we'll miss events, lacking any form of caching.
Now with a test.
Part of #3280.