Fix the proxy’s handling of WebSocket connections when subscribing to contract events and node data using the `eth_subscribe` RPC request.
Previously, the client connection that the subscription data was sent to was often in a closed state. It was determined that this connection was the wrong connection to forward the data in the first place. The connection was in fact generally the connection created for the Ethereum service check which was then (correctly and subsequently) closed after it had finished its operation.
The flow of a proxy request handling a WebSocket “eth_subscribe” RPC request is now as follows:
1. A WebSocket RPC request `”eth_subscribe”` is sent from a client to the proxy.
2. Proxy forwards the request to the node by way of a new instance of `RequestManager`.
3. When the node receives an event matching the subscription, it sends the event data back to same socket connection it received the request on (ie the specific instance of `RequestManager`).
4. The `RequestManager` fires the `”data”` event containing the subscription data, and this event is picked up in the proxy.
5. The proxy then forwards the subscription data on to the originating WS client connection.
All other requests (ie non-WS or WS RPC requests that are not `eth_subscribe`) will be serviced to/from the node using a single `RequestManager` instance.
Co-authored-by: Pascal Precht <pascal.precht@gmail.com>
Running embark's `blockchain` command resulted in a runtime error where the `blockchain`
module couldn't be found. This is a bug introduced in ed0d3afb4f where
we forgot to update `blockchainStackComponents` in Embark's engine accordingly.
Fixing this results in `embark blockchain` hanging. This is because there's a similar bug
in `blockchainStackCopmnonents` introduced in 3b8f8f9ea7.
This commit fixes both bugs by ensuring `embark-blockchain` and `embark-blockchain-client`
packages are loaded using the correct APIs.
If the version in the embark package's own `package.json` has a prerelease
identifier then appending `.x` to the major version isn't viable for resolving
the latest version of the template package that's in the same prerelease line;
a more complex semver range must be used:
```
"${pkg}@^${major}.${minor}.${patch}- <${major}.${minor}.${patch}"
```
When making use of the `useBuiltIns: 'usage'` option for @babel/preset-env
(which is the case for all transpiled packages in Embark's monorepo) a package
needs to have both @babel/runtime-corejs3 and core-js@3 specified as
dependencies.
Embark relies on certain specific plugin properties, e.g. registered compilers,
and retrieves them using `plugins.getPluginProperties('compilers', 'compilers')`.
In order to make this work in the testing environment, we need those same APIs
in the `embark-testing` package as well.
This commit adds necessary APIs to `Plugins` and `Plugin` to make registering and
loading compiler plugins work.
* build(deps): move deps needed by embark-basic-pipeline from packages/embark
Introduce additional refactors to ensure the packages can be resolved by the
basic pipeline's webpack child process.
* build(deps): move @types/os-locale from packages/embark to packages/core/i18n
* build(deps) move @types/globule from packages/embark to packages/plugins/coverage
* build(deps): refactor stack/{api,proxy,webserver} deps relative to packages/embark
* build(deps): remove unneeded @types/async dep from packages/stack/test-runner
* build(deps): remove unneeded deps from packages/embark
* build(deps): upgrade create-react-app for cockpit by bumping react-scripts to latest
Also get rid of a peer dependency warning related to storybook. After some
investigation it seems that storybook can't practically (at present) be made
aware of CRA in the same project satisfying storybook's peer deps, so it's best
to just satisfy all of them explicitly, which in any case won't interfere with
CRA (react-scripts).
* build(deps): move @types/os-locale from packages/embark to packages/core/i18n
* build(deps) move @types/globule from packages/embark to packages/plugins/coverage
* build(deps): refactor stack/{api,proxy,webserver} deps relative to packages/embark
* build(deps): remove unneeded @types/async dep from packages/stack/test-runner
* build(deps): remove unneeded deps from packages/embark
Introduce some light refactoring related to embark-testing facilities, and also
a configurable stdout option so the output of the reporter implemented in this
package isn't confusingly mixed into unit test reporting for this package.
BREAKING CHANGE:
node: >=10.17.0 <12.0.0
npm: >=6.11.3
yarn: >=1.19.1
node v10.17.0 is the latest in the 10.x series and is still in the Active LTS
lifecycle. Embark is still not compatible with node's 12.x and 13.x
series (because of some dependencies), otherwise it would probably make sense
to bump our minimum supported node version all the way to the most recent 12.x
release.
npm v6.11.3 is the version that's bundled with node v10.17.0.
yarn v1.19.1 is the most recent version as of the time node v10.17.0 was
released.
* refactor(@embark/dapps/tests/app): use function syntax
These changes don't fix the race conditions related to the test dapp's tests
but are a step in the right direction.
* refactor(@embark/dapps/tests/contracts): adjustments to get tests passing
Further refactoring is needed re: potentially duplicated or overlapping logic
in `packages/plugins/solidity-tests` and
`packages/core/utils/src/solidity/remapImports.ts`, as well in disabled test
dapp tests
* test(dapps/tests/app): temporarily disable intermittently failing tests
They are failing because of a race condition; once that race condition has been
fixed these tests should be reenabled.
* fix(@embark/solidity-tests): fix importing the library for the tests
* build(@embark/stack/blockchain-client): remove unneeded typescript related scripts and deps
In addition to being unneeded their presence is causing build errors.
* build(embark): remove unneeded typescript related scripts and deps
In addition to being unneeded their presence is causing build errors.
* build(@embark/plugins/ethereum-blockchain-client): remove unneeded typescript related scripts, deps
In addition to being unneeded their presence is causing build errors.
* build(@embark/plugins/ganache): remove unneeded typescript related scripts and deps
In addition to being unneeded their presence is causing build errors.
* build(@embark/plugins/geth): remove unneeded typescript related scripts and deps
In addition to being unneeded their presence is causing build errors.
* fix(@embark/plugins/transaction-logger): require 'web3' not 'Web3'
* fix(@embark/utils/solo): spawn npm(.cmd) instead of npx(.cmd)
* test(@embark/plugins/basic-pipeline): add test stub
* test(@embark/stack/blockchain): add test stub
Fix a lot of bugs and reenable a couple of modules
Some tests were kept disabled, mostly the ENS and EmbarkJS tests
Those need to add back a fairly significant feature to work
Add back missing solidity contracts
In d6bf5c24b9 we've ensured that certain modules of
embark only executed if their functionality is actually enabled.
This broke one of our tests in the communication module.
This commit fixes the test by explicitly enabling the module's functionality.
* fix: fix tests hanging because the console is not started
* fix(@embark/proxy): send back errors correctly to the client
Code originally by @emizzle and fixed by me
* feat(@embark/test-runner): add assert.reverts to test reverts
* fix: make test app actually run its test and not hang
* fix(@embark/proxy): fix listening to contract event in the proxy
* feat(@embark/test-runner): add assertion for events being triggered
* docs(@embark/site): add docs for the new assert functions
* feat(@embark/test-runner): add increaseTime util function to globals
* docs(@embark/site): add docs for increaseTime
Contains bug fixes to get parity to work as a blockchain client.
**NOTE:** Please merge https://github.com/embark-framework/embark/pull/1950 first before merging this PR, as this PR contains removal of dependencies from `packages/embark` that are needed for geth. So if this is merged first, and the geth PR (https://github.com/embark-framework/embark/pull/1950) is not merged, then geth will not have some of the dependencies it needs.
**NOTE:** Running a dapp with `client: “vm”` does not work, however I am not sure if that is intentional in v5 or not.
**NOTE:** Running `embark test` in the demo DApp does work, although there is an error that I am assuming does actually pertain to this PR:
```
made request without listener: whisper:node:register
Trace:
at EmbarkEmitter.trace [as request] (/Users/emizzle/Code/__Github/embk-fw/embark/packages/embark/src/lib/core/events.js:142:13)
at new request (/Users/emizzle/Code/__Github/embk-fw/embark/packages/embark/src/lib/modules/geth/index.js:49:17)
at Plugin.loadInternalPlugin (/Users/emizzle/Code/__Github/embk-fw/embark/packages/embark/src/lib/core/plugin.js:117:10)
at Plugins.loadInternalPlugin (/Users/emizzle/Code/__Github/embk-fw/embark/packages/embark/src/lib/core/plugins.js:96:40)
at Engine.loadInternalPlugin [as registerModule] (/Users/emizzle/Code/__Github/embk-fw/embark/packages/embark/src/lib/core/engine.js:69:18)
at Engine.registerModule [as blockchainComponents] (/Users/emizzle/Code/__Github/embk-fw/embark/packages/embark/src/lib/core/engine.js:183:10)
at Engine.apply [as registerModuleGroup] (/Users/emizzle/Code/__Github/embk-fw/embark/packages/embark/src/lib/core/engine.js:104:18)
at registerModuleGroup (/Users/emizzle/Code/__Github/embk-fw/embark/packages/embark/src/cmd/cmd_controller.js:740:16)
at nextTask (/Users/emizzle/Code/__Github/embk-fw/embark/node_modules/async/dist/async.js:5324:14)
at next (/Users/emizzle/Code/__Github/embk-fw/embark/node_modules/async/dist/async.js:5331:9)
at /Users/emizzle/Code/__Github/embk-fw/embark/node_modules/async/dist/async.js:969:16
at callback (/Users/emizzle/Code/__Github/embk-fw/embark/packages/embark/src/lib/core/engine.js:47:9)
at Client.done (/Users/emizzle/Code/__Github/embk-fw/embark/packages/core/core/src/ipc.js:46:11)
at Client.emit (/Users/emizzle/Code/__Github/embk-fw/embark/node_modules/event-pubsub/es5.js:74:21)
at Socket.connectionClosed (/Users/emizzle/Code/__Github/embk-fw/embark/node_modules/node-ipc/dao/client.js:201:20)
at Socket.emit (events.js:198:13)
```
**TODO:** The `embark eject-webpack` command needs to be updated. @michaelsbradleyjr - do you have suggestions as to how we could port this over? We could **assume** that the `basic-pipeline` plugin will be installed, and then read the embark config and overrides from the file system, however this feels like we are sort adding a dependency to a plugin, which is not right. Any suggestions?
* bugfix(@embark/embark): fix default values so choosing different environment doesn't cause hang
* bugfix(@embark/stack/watcher): add callback so initializating watcher does not hang cmd controller
Change components/TexEditor re: how monaco-editor is loaded, since difficulties
with monaco-editor were the original motivation for "ejecting" the CRA setup.
Specify "nohoist" for `embark-ui/react-scripts` and all its dependencies to
avoid potential headaches when upgrading the react-scripts, i.e. upgrading to
newer versions of create-react-app.
It's principal use was in `stack/compiler`, but that package was refactored
in #1878 to use async/await. There was a replica in
`packages/embark/src/lib/utils/async_extend.js` but it wasn't being used to do
async operations and could be replaced with `Object.entries().forEach()`. It
was required by `plugins/vyper` but the wrapper's custom `eachObject` method
wasn't being used. `stack/api` had it as a dependency but wasn't using it.
* chore: test framework
* chore(@embark/testing): introduce Plugin apis and other changes
* refactor(@embark/deployment): use new testing APIs for tests
* feat(@embark/compiler): support :before and :after hooks on event compiler:contracts:compile
* style: keep root package.json sorted cleanly
* style: formatting with prettier where print-width is 200
Per the suggestion of @iurimatias
* WIP: PR review
* WIP: PR review
* WIP: refactor to check if file.path is within dappPath() or os.tmpdir()
* WIP: PR review
* WIP: use native async functions (and related JS lang constructs) instead of the async library
* WIP: async.eachObject is no longer a dependency
* fix: fix test-app, contracts index file and reload on change
* fix(@embark/cmd_controller): fix missing nodes
Was removed by accident
* feat(@embark/demo): add favicon to embark demo
* chore(@embark/cockpit): change favicon to new Embark logo
* fix(@embark/embarkjs-ens): fix ENS config for embarkjs-ens
* remove comments
When an options object is not supplied the tests fail because the constructor
expects the options argument to be an object, i.e. referencing
`options.plugins` should not cause a TypeError
* chore(@embark/): move embarkjs packages to their own folder
* chore(@embark/): rename embark-ui folder to cockpit
* chore(@embark/): rename packages already in logical folders to remove embark- prefix
chore(@embark/): rename packages already in logical folders to remove embark- prefix
update package.json files to use correct eslint config
remove core/* from package.json workspaces
* fix(@embark/cmd_controller): fix embark blockchain command
* feat(@embark/blockchain): ping endpoint before starting the node
also use web3 instead of function to ping endpoint
* dont override console.error completely
* fix embarkjs generation
fix ens setProvider
fix embarkjs objects
fix generated embarkjs provider
generate contracts
fix embarkjs-ens
* address some of the issues in the code review
* feat(@embark/specialconfigs): introduce dynamic addresses
This commit introduces a new Smart Contract configuration addressHandler
that lets users define a function to "lazily" compute the address of the
Smart Contract in question.
This is useful when a third-party takes care of deploying a dependency
Smart Contract, but the address of that Smart Contract only being available
at run-time.
Example:
```
deploy: {
SimpleStorage: {
fromIndex: 0,
args: [100],
},
OtherContract: {
deps: ['SimpleStorage'],
address: async (deps) => {
// use `deps.contracts.SimpleStorage` to determine address
},
abiDefinition: ABI
},
}
```
In the example above, OtherContract will be deployed after SimpleStorage
because it uses the deps property to define Smart Contracts that it depends
on. All dependencies are exposed on the addressHandler function parameter
similar to deployment hooks.
Closes#1690
There was an error that would display on the second+ run of embark, that was causing by trying to read JSON of an empty file.
The solution was a combination of ensuring the file existed with defaults when enabled, and also ensuring we await the saving of the file.
Included is a bit of a refactor of how the tracking functions handled the “current chain”. Hopefully, this should make things more clear.
Tests have been updated accordingly.
* fix embarkjs generation
fix ens setProvider
fix embarkjs objects
fix generated embarkjs provider
generate contracts
fix embarkjs-ens
* address some of the issues in the code review
* address some of the issues in the code review
* address some of the issues in the code review
* address some of the issues in the code review
* refactor(@embark/cmd-controller): re add cargo for file watcher
(cherry picked from commit 5c77b4000742a14f42a6dc5c145672b9cd0957a3)
* use generate:all instea dof the event
This is crucial as the storage module tries to generate/eval artifacts that
depend on a `web3` object inside the VM scope. The registration of that object
can only be garuanteed when the `blockchain:started` event has been fired.
This commit ensures these modules are not spinned up in parallel as we're running
into race conditions.
DApps generated with `new --simple` don't have a blockchain config so ensure
that the blockchain proxy and listener (dev_txs) use appropriate provider
settings when setting up web3.
Without these changes, in a fresh `embark new --simple` project the `embark
blockchain` command will encounter unhandled promise rejection with respect to
the proxy; `embark run` will encounter similar with respect to dev_txs.
Previously, when using "development embark" (i.e. `embark` command in the
monorepo) the template generator would always use `yarn install` to install a
template's dependencies. However, if a template relies on `package-lock.json`,
as is the case for `embark-create-react-dapp-template`, that behvaior can cause
serious problems. So, have the choice of install command depend not only on
whether we're using "development embark" but also whether the template has a
`package-lock.json` file.
NOTE: this change can probably result in problems given that `yarn install` and
`npm install` behave differently re: symlinks in `node_modules`. However, such
problems would never show up with production installs of the `embark`
packaage. Moreover, after dropping `embarkjs-connector-web3` we don't presently
have a situation where a monorepo package would be `yarn link`ed into an
instantiated template's `node_modules`. So, in effect, the changes introduced
in this PR allow existing templates to work correctly whether using production
or monorepo embark. The impending embark v5 refactor should deprecate `embark
new` in favor of the yet unfinished `embark init`.
When passing the name of a build pipeline that doesn't exist, e.g.
`embark run|build --pipeline foo`, `writeStats` would cause an unhandled
promise rejection because its config argument is falsy. Checking for that and
returning early from `writeStats` allows the `"no webpack config has the name
'foo'"` error to be reported correctly.
The ignore paths for the default webpack config and babel-loader-overrides
helper script weren't properly adjusted after the legacy pipeline was split-out
into its own package in the monorepo.
Upgrade all dependencies on web3/web3-* v1.0.0-beta.37 to v1.2.1.
Make various adjustments related to the previous convention of
`"web3": "1.0.0-beta"` in `embark.json` signifying that embark's own web3
dependency should be used in dapp builds.
Fix bugs in library manager, including a switch from using the
live-plugin-manager package to using npm in a child process to install
`"versions"` dependencies specified in `embark.json` when a specified version
doesn't match up with embark's own version for that package.
Avoid race conditions when installing `"versions"` by completing all installs
prior to starting other services. If an install fails, then after all the
installs have completed or failed the embark command will exit with error.
Change various comments and update docs to reflect the new default of web3
v1.2.1.
In https://github.com/embark-framework/embark/commit/87d92b6091 we've introduced a feature in our test runner
to report exact gas costs used when calling Smart Contract methods that
cause computation.
Embark keeps a list of deployed contracts in memory and for all transactions
happening during tests, it'll match them to the contracts that performed them
to produce the report logs.
Unfortunately, we've been [resetting that memory](https://github.com/embark-framework/embark/commit/87d92b6091#diff-92b4f79a0473160fe700440b1ced5204R140) of deployed contracts
after every test, making it practically impossible for Embark to find
matching contracts for any transactions, because contracts aren't necessarily
redeployed per spec, resulting in no additional transaction logs whatsoever.
This commit ensures that the memory is never erased and at the same time
ensures it's not leaking infinitely in case multiple contracts are
deployed multiple times in the same process over several specs.
We've introduced a regression in 536a4029ba where the default
`value` sent to payable Smart Contract methods is `-1`, resulting in errors as it's not
a valid value.
For payable methods inside of the contract interatction section of cockpit, the value input has been updated to pass wei to the API. The input field now accepts value like `100 ether` or 25 szabo`. The value entered is automatically converted to wei and shown to the user in real time.
Additionally, the input is validated for a correct value, with an error shown to the user for incorrectly entered values.
A tooltip has been added to help the user enter correct values.
UI updates can be seen in video here: https://monosnap.com/file/642cHH2HxDeiFLzB2VLqHP9GuqoRfz
This is due to a bug that has been introduced in d10c0b7951 where we end up
sending values as numbers to web3's `toWei()` utility. The `value` is always sent as number
and in fact always defined, so we just need to check for it's type and convert it to a string
according to `toWei()`'s API.
When running tests against non-simulated blockchain nodes, even for simplex
Smart Contracts, deployment transactions would exceed the block gas limit.
E.g. running `embark blockchain` in one process and `embark test --node embark`
in another, inside our demo application, will throw an error when Embark attempts
to deploy its `SimpleStorage`:
```
Compiling contracts
Compilation done
[SimpleStorage]: error deploying =SimpleStorage= due to error: Returned error: exceeds block gas limit
Error deploying contracts. Please fix errors to continue.
Error deploying contracts. Please fix errors to continue.
terminating due to error
Error deploying contracts. Please fix errors to continue.
```
The reason for that is because in https://github.com/embark-framework/embark/pull/1650, we've introduced a static
gas estimation for Smart Contract deployment that is just right below Ganache's
maximum gas limit of `6721975`, since Ganache tends to underestimate gas for
complex Smart Contracts due to its [low base fee](8ad1ab29de/lib/utils/gasEstimation.js (L33-L39)).
The static gas estimation would apply any time we're in a test context, but we
didn't take into account the case where tests are executed against nodes
other than the simulated environment.
As mentioned in the comments in the linked PR:
> If this is not spec'ed at all, I wonder what complications it could cause when
> at some point we maybe switch to not using Ganache anymore for tests, or even
> the user itself (which I think is a reasonable thing to do).
This causes the error described above because we easily reach the block gas limit
with just two Smart Contracts and Embark already deploys a few Smart Contracts for
ENS.
So basically what we want is to use the static gas estimation when we know
the node we're connecting to is Ganache. In all other cases we can rely on the
standardized gas estimation offered by the node.
Adjust the API endpoints to augment transaction objects with a timestamp
property from their corresponding blocks. This removes the necessity to copy
the timestamp property from a block to its transactions in the client.
Introduce a `formatTimestampForDisplay` util function in Cockpit for
consistently transforming timestamps into relative or absolute dates, depending
on whether a date is sometime during the last day.
The property is set/available on EmbarkJS in the artifacts, i.e. in a browser
build; but sometime since v3.2.4 it was no longer availble in the cli dashboard
environment.
`EmbarkJS.Messages.isAvailable()` in some cases return synchronously (when whisper isn't
set up), in other cases asynchronously. This actually breaks our demo application for
the following reason:
We check for Whisper's availability via:
```
EmbarkJS.Messages.Providers.whisper.getWhisperVersion((err, _version) => {
if (err) {
return console.log(err);
}
this.setState({whisperEnabled: true});
});
```
There's a couple of problems here:
- This code will break right away when whisper isn't available, resulting in an error:
```
Cannot read property _requestManager of undefined
```
- The reason this error happens is because there's no `web3` object available inside
our EmbarkJS.Messages code. Even though there **is** a web3 object, EmbarkJS.Messages
doesn't know about this because it only sets it when its `setProvider()` API is called,
which effectively doesn't happen at all when Whisper isn't enabled on the connected
node
- While this could be fixed with a simple check on whether EmbarkJS.Messages' internal
`web3` references is a thing, really what should be used in the demo is the `isAvailable()`
API.
`isAvailable()` should always return a promise (similar to `EmbarkJS.Storage.isAvailable()`.
This commit ensures that `isAvailable()` always returns a promise and changes the demo
template to use `isAvailable()` over `getWhisperVersion()`.
Effectively deprecate the `embarkjs-connector-web3` package but don't introduce
a breaking change by simply not loading the plugin if it's specified in a
DApp's `embark.json`. If the deprecated plugin is specified, display a message
indicating the plugin was ignored and suggesting it be removed from the
project's `embark.json` and `package.json`.
This PR adds a command to get full account details from the contradts config that includes info like private key.
The existing, and similar command, `blockchain:provider:contract:accounts:get` would only return account addresses if they existed, and not the full account info.
This check was already made when sending messages to whisper channel, however, we didn't
perform the same check when subscribing to channels within cockpit.
The webserver's job is to serve files generated by Embark's built-in pipeline,
however, since v4 users can choose they front-end tool to take care of building,
bundling and packing their DApps. Usually these tools come with a built-in
dev server as well.
Therefore, when the pipeline is turned off (which also soon will be the default),
there's not need start a webserver.
In https://github.com/embark-framework/embark/pull/1119 we've introduced a feature where
users can provide an already compiled ABI for Smart Contracts so they can be used
without the need to compiling them again.
This also means that, internally, those Smart Contract object won't have any
bytecode attached to it.
Later on, in 387d33a076, we've introduced a warning
which is rendered when a Smart Contract's bytecode is too large. The check expects
a Smart Contract object to have bytecode associated to it, which will break the code
in cases where a Smart Contract has already an ABI and therefore didn't need compilation.
This commit ensures we only check a Smart Contract's bytecode when bytecode exists for
the Smart Contract in question.
When trying to either sending, or listening to whisper channels within
Embark's consoles (CLI and Cockpit), the console outputs an error that
`EmbarkJS` isn't defined.
E.g. running:
```
> EmbarkJS.Messages.listenTo({ topic: ['somechannel'])
```
Outputs:
```
EmbarkJS is not defined
```
This seemed very odd as outputting `EmbarkJS` and all of its members
worked totally fine.
It turns out that both methods, `listenTo()` and `sendMessage()`, in
`EmbarkJS.Messages` are not necessarily what one thinks they are.
EmbarkJS decorates both methods to create some options that need to be
available in the delegate, two of them being `EmbarkJS.Utils.toAscii`
and `EmbarkJS.Utils.fromAscii` (ac76a40a61/packages/embarkjs-whisper/src/index.js (L43-L62) and ac76a40a61/packages/embarkjs-whisper/src/index.js (L64-L73))
These two global references to `EmbarkJS` are the only ones in `embarkjs-whisper`
and they are not the same reference as the `EmbarkJS` sandbox global
registered in the VM here ac76a40a61/packages/embark-code-runner/src/index.ts (L33).
In other words, the `EmbarkJS is not defined` message actually refers
to the `EmbarkJS` in the wrapping method, not the one registered with
the VM, which is also why inspecting the `EmbarkJS` object through the
VM works fine.
Since the `toAscii()` and `fromAscii()` methods in use are really just
facades around `web3.utils.[fromAscii|toAscii]()`, we can replace the
global EmbarkJS dependency with web3 utils that are already available
anyways.
Upgrade chokidar to a version that's compatible with NodeJS v12.x.
Unfortunately, embark has other transitive dependencies that are not compatible
with v12.x, but upgrading chokidar is still a good step.
Iuri Matias
emizzle
Jonathan Rainville
Pascal Precht
Michael Bradley, Jr
dependabot-preview[bot]
André Medeiros