`Engine.startEngine(cb)` potentially exits the current process if
any registered plugins emits an error in the bootstrap phase.
This limits consumers of this API to react accordingly. Embark's APIs
should be considered library APIs that propagate errors, but let callers
decide what to do with them.
This commit ensures we keep propagating the error of `startEngine()` without
exiting the current process.
Client traffic with the communication provider node, e.g. a whisper node, is
not proxied so make the default port 8547 instead of 8557. It's not a technical
problem for it to be 8557, but our convention to present has been for 855*
ports to be proxied while the upstream is an 854* port.
Update the boilerplate and demo templates to match.
1. get rid of the whisper layer and handle everything in the communication layer
2. Create a gethWhisper and a parityWhisper plugin that has the same files as packages/embark/src/lib/modules/geth, except with a whisperClient instead of gethClient, and will include most of https://github.com/embark-framework/embark/blob/master/packages/plugins/whisper/src/index.js.
3. Get rid of any whisper registration in geth.
4. In the whisperGeth and whisperParity plugins, modify the request for communication:node:register, to end up calling `startWhisperNode`
When using websockets, the simulator was starting on port `8545` instead of `8546`, so the blockchain check was failing.
Update the simulator port to be websockets port from the config when using websockets.
When whisper was being started in a new process, the message displayed in the console was “Starting Blockchain node in another process”. This is misleading as it is the same message that geth/parity use to start up their processes.
Update whisper initialisation process messaging so it states “Starting Whisper node in another process”.
Improve the reliability of the expiration unit test in the test dapp by explicitly setting the `block.timestamp` and comparing an expiration value against that.
This improves on the current implementation that relies on time passed which varies depending on the speed of unit tests run (CPU speed, logs shown, etc).
When `eth_unsubscribe` is received in the proxy, ensure this request is forwarded through on the correct socket (the same socket that was used for the corresponding `eth_subscribe`).
Move subscription handling for `eth_subscribe` and `eth_unsubscribe` to RpcModifiers (in `rpc-manager` package).
For each `eth_subscribe` request, a new `RequestManager` is created. Since the endpoint property on the proxy class was updated to be a provider, the same provider was being assigned to each new `RequestManager` and thus creating multiple event handlers for each subscription created. To circumvent this, we are now creating a new provider for each `RequestManager`.
Co-authored-by: Pascal Precht <pascal.precht@googlemail.com>
This PR adds support for EIP-712 (https://github.com/ethereum/EIPs/blob/master/EIPS/eip-712.md), by allowing the signing of typed data transactions using the `eth_signTypeData_v3` or `eth_signTypedData` request.
Add a module called `embark-rpc-manager` to handle all RPC transaction modifications.
The internal events management module extracted to `embark-core`. This allows other modules (ie plugins) to use a private instance of Embark’s event API.
Remove transaction modifications in `embark-accounts-manager` and leave only the functionality in this module that handle account management (ie funding of accounts). Other functionality was moved to `embark-rpc-manager`.
- Transactions should now reflect when a new node account is added via `personal_newAccount` or via the node itself (and once `eth_accounts` is requested).
- In the proxy, errors are handled for all cases and are now JSON-RPC spec-compliant.
- Always register `eth_signTypedData` RPC response and display error message if account cannot be found.
NOTE: Updated yarn.lock due to conflict after rebase.
Avoid race condition with rpc modifications
Refactor proxy to make request handling asynchronous, and to allow circumvention of forwarding requests to the node in cases where it is not needed.
Move loading of plugins to after core and stack components (in cmd_controller).
We should load custom plugins LAST (ie after all core and stack components, and after all Embark plugins). To do this, we disable automatic loading of plugins in the config, and allow them to be loading explicitly in cmd_controller.
Create a system that allows registration of custom plugins. Each plugin is generated in EmbarkJS by instantiating the registered class. For example, we can register a Plasma plugin and it will be generated in to EmbarkJS.Plasma with an instantiation of `embarkjs-plasma`. The DApp itself can then do what it wants with `EmbarkJS.Plasma` (ie call `EmbarkJS.Plasma.init()`.
NOTE: loading of custom plugins needs to be applied to all other commands in `cmd_controller`. This PR only loads custom plugins for the `run` command.
NOTE: This PR is based on branch fix/ws-eth-subscribe (#1850). I will rebase this PR when (and if) #1850 is merged.
In 776db1b7f7 (diff-5cab125016e6d753f03b6cd0241d5ebbR267) we've introduced the ability to add
a `priority` parameter to plugin actions, so the order of actions can be semi-ensured.
That commit also introduced typings for that new API but it actually didn't match the
implementation of the API, namely that the second parameter of `Plugin.registerActionForEvent()`
can be either an options object **or** a callback function.
This forces consumers to call the API as `registerActoniForEvent(name, undefined|null, callback)`
which shouldn't be necessary.
BREAKING CHANGE:
Related to #1985. Prior to Embark's minimum supported version of Node.js being
bumped to to 10.17.0, Embark was incompatible with any relatively recent
release of the `ipfs-http-client` package.
While *internal* changes are needed re: ipfs's `Buffer` export for
e.g. `embark_demo` to function correctly *(and this PR makes those changes)*,
Embark otherwise runs/tests okay.
Keep in mind #2033.
However, if a dApp author were to explicitly `require('ifps-api')` in the
front-end that wouldn't work as before; and swapping `"ipfs-http-client"` for
`"ipfs-api"` might also not be enough — there are API changes that dApp authors
would need to consider, regardless of Embark presently supplying the dependency
and EmbarkJS wrapping around it.
Closes#1994.
Hook `embark-profiler` into the engine's contracts group.
Adjust `request` signatures to match v5 usage.
Ensure gas numbers are formatted consistently in the ascii table.
Display non-json output in cockpit's console as monospace/pre to preserve
profiler table formatting.
Embark's `File.content` is an asynchrononus getter that potentially downloads the
contents of a file from its `externalUrl`. This can potentially fail but is not properly
reported to the user as the error itself is ignored.
This commit ensures errors are propagated and displayed to the user.
The blockchain module should not contain any ethereum-specific code, and currently it contains a check to see if the blockchain has already been started.
This PR moves this check to blockchain client (ie geth or parity). This check function is registered along with the started callback.
style(@embark/stack/blockchain): add missing semicolon
Turns out that, when Embark tries to replace ENS subdomains, it would fail
if `subdomains` aren't defined in the namesystem configuration.
This commit adds a safeguard so that not defining `subdomains` is fine.
This commit fixes the issue that it wasn't possible anymore to use named constructor arguments
in Smart Contract configurations.
For example, the following Smart Contract expects two constructor arguments:
```solidity
contract SomeContract {
constructor(address[] _addresses, int initialValue) public {}
}
```
The first being a list of addresses, the second one a number. This can be configured as:
```js
SomeContract: {
args: [
["$MyToken2", "$SimpleStorage"],
123
]
}
```
Notice how the order of arguments matters here. `_addresses` come first in the constructor,
so they have to be defined first in the configuration as well.
Another way to configure this is using named arguments, which is what's broken prior to this commit:
```js
SomeContract: {
args: {
initialValue: 123,
_addresses: ["$MyToken2", "$SimpleStorage"]
}
}
```
Using a notation like this ^ the order no longer matters as Embark will figure out the right
values for the constructor arguments by their names.
The reason this is broken is because there are several modules in Embark that register and
run a `deployment:contract:beforeDeploy` action, which are allowed to mutate this configuration
object. One of those modules is the `ens` module, which searches for ENS names in the arguments
and figure out whether it has to replace it with a resolved address.
One thing that particular module didn't take into account is that `args` could be either and
array, or an object and will always return an array, changing the shape of `args` in case it was
an object.
This is a problem because another module, `ethereum-blockchain-client`, another action is registered
that takes this mutated object in `determineArguments()` and ensure that, if `args` is actually an
object, the values are put in the correct position matching the constructor of the Smart Contract in
question.
One way to solve this was to use the newly introduced [priority](https://github.com/embark-framework/embark/pull/2031) and ensure
that `ens` action is executed after `ethereum-blockchain-client`'s.
However, the actual bug here is that the ENS module changes the shape of `args` in the first place,
so this commit ensures that it preserves it.
Michael Bradley, Jr
dependabot-preview[bot]
Pascal Precht
emizzle
Jonathan Rainville
Iuri Matias