When testing rollback feature in the Cosmos SDK, we found that the app hash
in Tendermint after rollback was the value after the latest block, rather than
before it.
Co-authored-by: Callum Waters <cmwaters19@gmail.com>
While I'd hoped to be able to make the socket client less weird, I
think that this is a nice middle ground in terms of improving
readability and removing the vestigal components without breaking
anything or radically changing the underlying assumptions.
In the future we'd want to have requests be identified by a request
ID, and then we could drop the request tracking logic in the client
entirely, and this is protocol breaking. The alternatives aren't
substantively different than the current implementation.
This follows along in the spirit of #7845 but is orthogonal to
removing `CheckTxAsync` (which will come after the previous commit
lands,) so I thought I'd get it out there earlier.
After poking around #7828, I saw the oppertunity for this cleanup,
which I think is both reasonable on its own, and quite low impact, and
removes the math around process start time.
Now that shutdown is handled by contexts in most cases, I think it's
fair to cleanup the way this reactor shuts down. Additionaly there
were a few cases where the `blockSyncOutBridgeCh` was misshandled and
could have lead to a deadlock which I observed in some tests
This change adds logic to double the message delay bound after every 10 rounds. Alternatives to this somewhat magic number were discussed. Specifically, whether or not to make '10' modifiable as a parameter was discussed. Since this behavior only exists to ensure liveness in the case that these values were poorly chosen to begin with, a method to configure this value was not created. Chains that notice many 'untimely' rounds per the [relevant metric](https://github.com/tendermint/tendermint/pull/7709) are expected to take action to increase the configured message delay to more accurately match the conditions of the network.
closes: https://github.com/tendermint/spec/issues/371
* evidence: Refactored the evidence message to send/recieve Evidence instead of EvidenceList
Peers send evidences one by one. The refactored code reflects this behaviour.
* Update internal/evidence/reactor.go
Co-authored-by: M. J. Fromberger <fromberger@interchain.io>
Co-authored-by: M. J. Fromberger <fromberger@interchain.io>
Co-authored-by: William Banfield <4561443+williambanfield@users.noreply.github.com>
This is clearly a cob-web in the code, and may predict a solution to #7729, though this is difficult to backport because we don't have contexts in 0.35
Responses are constructed from requests using MakeResponse, MakeError, and
MakeErrorf. This ensures the response is always paired with the correct ID,
makes cases where there is no ID more explicit at the usage site, and
consolidates the handling of error introspection across transports.
The logic for unpacking errors and assigning JSON-RPC response types was
previously duplicated in three places. Consolidate it in the types package for
the RPC subsystem.
* update test cases
This pull request merges in the changes for implementing Proposer-based timestamps into `master`. The power was primarily being done in the `wb/proposer-based-timestamps` branch, with changes being merged into that branch during development. This pull request represents an amalgamation of the changes made into that development branch. All of the changes that were placed into that branch have been cleanly rebased on top of the latest `master`. The changes compile and the tests pass insofar as our tests in general pass.
### Note To Reviewers
These changes have been extensively reviewed during development. There is not much new here. In the interest of making effective use of time, I would recommend against trying to perform a complete audit of the changes presented and instead examine for mistakes that may have occurred during the process of rebasing the changes. I gave the complete change set a first pass for any issues, but additional eyes would be very appreciated.
In sum, this change set does the following:
closes#6942
merges in #6849
This is the interface shared by types that can be used as event data in, for
example, subscriptions via the RPC.
To be compatible with the RPC service, data need to support JSON encoding.
Require this as part of the interface.
Remove the pubsub.Query interface and instead use the concrete query type.
Nothing uses any other implementation but pubsub/query.
* query: remove the error from the Matches method
* Update all usage.
There are no further uses of this package anywhere in Tendermint.
All the uses in the Cosmos SDK are for types that now work correctly with the
standard encoding/json package.
The main change here is to use encoding/json to encode and decode RPC
parameters, rather than the custom tmjson package. This includes:
- Update the HTTP POST handler parameter handling.
- Add field tags to 64-bit integer types to get string encoding (to match amino/tmjson).
- Add marshalers to struct types that mention interfaces.
- Inject wrappers to decode interface arguments in RPC handlers.
The problem with the `TestStateFullRound1` is that the state that we are observeing, `cs`, can advance to the next height before we query its data. Specifically, on line `388`, when we called `validatePrevote`, the `cs` State had already advanced to height 2, so querying that State for the votes of height 1 either yielded nil or an erroneous value. This change adds a `ensurePrevoteMatch` function that checks that the prevote occurred and checks that it is for the expected block at the same time. If this change looks reasonable I can just apply the same fix to all of the places where we perform `ensurePrevote` followed by `validatePrevote` to use this function instead.
