You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hotfix Checkpoint handling of API server projections (#2152)
The projections in the API server were not correctly handing
`Checkpoint` events (introduced with event log rotation).
I tried to create a regression test, but we don't have a `Gen
[StateChanged tx]` generators that honors the constraints required for
`aggregate` so I could not get hold of an expected list of
`StateChanged` events with a consistent `Checkpoint` to do property
testing.
The whole `Projection` business is a bit contrived since we have the
full `StateChanged` event stream available (the API server is now an
`EventSink`). As we are also not needing to track changes onto the
projected values (resource-specific websockets / subscriptions would
require that), we should consider dropping the whole mechanism and just
stick with a `getHeadState :: STM (HeadState tx)` + getter functions (or
lenses) to acquire the relevant things in the HTTP and WS API.
This PR though is a hotfix and I kept the diff to a minimum (especially
as its not covered by tests!)
---
* [x] CHANGELOG updated
* [x] Documentation update not needed
* [x] Haddocks updated
* [ ] No new TODOs introduced
- An XXX note on the same as above
Copy file name to clipboardExpand all lines: CHANGELOG.md
+24Lines changed: 24 additions & 0 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -10,8 +10,32 @@ changes.
10
10
11
11
## [0.23.0] - UNRELEASED
12
12
13
+
## [0.22.4] - UNRELEASED
14
+
13
15
- Accept additional field `amount` when depositing to specify the amount of Lovelace that should be depositted to a Head returning any leftover to the user.
14
16
17
+
- Fix API not correctly handling event log rotation. This was evident in not
18
+
being able to use `/commit` although the head is initializing or outdated
19
+
information in the `Greetings` message.
20
+
21
+
- Ignore snapshot signatures of already confirmed snapshots. This was previously
22
+
resulting in the node waiting for the accompanying snapshot request and
23
+
occurred when running heads with mirror nodes.
24
+
25
+
- Fix an internal persistent queue blocking after restart when it reached
26
+
capacity.
27
+
28
+
- Handle failing lease keep alive in network component and avoid bursts in
29
+
heartbeating.
30
+
31
+
## [0.22.3] - 2025-07-21
32
+
33
+
* Change behavior of `Hydra.Network.Etcd` to fallback to earliest possible
34
+
revision if `last-known-revision` is missing or too old. This can happen if a
35
+
node is down for a long time and the `etcd` cluster compacted the last known
0 commit comments