When running unit tests inside a project that configures ENS subdomains using Smart Contract directives,
the tests will output an error because of that particular Smart Contract's `deployedAddress` being `undefined`.
This happens only when running tests, not when deploying the Smart Contracts including the custom ENS setup.
It turns out that Embark attempts to compile and deploy the Smart Contracts of the project *twice* - once
before tests are executed, and another time **after** tests are done executing.
Both compilations/deployments are triggered through our custom `config()` function within test context,
which ensures all Smart Contracts are deployed before tests are executed.
This explains a few things:
1. There's no such issue when running `embark run`, in fact the custom ENS subdomains work perfectly fine
2. That's also why the tests are passing fine as well as the first compilation/deployment doesn't have any issues.
The errors only appear *after* the tests have been executed.
Still, this raises a few more questions, mainly
- Why is the Smart Contract's `deployedAddress` property `undefined` when `config()` is executed a second time?
- Why is `config()` executed a second time in the first place?
Let's look into both of these.
Assuming that there's a valid reason that `config()` is called twice, it's remains unclear why that property
of a Smart Contract object is `undefined` in the second run. The reason for that is that Embark determines whether
or not a particular Smart Contract should be deployed. One of the routines ensures that only the Smart Contracts
configured for deployment are actually attempted to be deployed (as opposed to just deploying all Smart Contracts
found in the file system).
It turns out that the second `config()` call is done without any Smart Contract configuration, resulting basically in no deployment
for any Smart Contract of the application. The `address` and `deployedAddress` of a Smart Contract object are however only set
*after* deployment, resulting in them being `undefined`.
**All this makes sense.**
`config()` is simply not designed for being used with an empty `contracts: {}` config. This raises another question:
Why is `config()` called with an empty configuration? This leads us to the second point.
It does seem a bit weird that Embark tries to configure compile *and* deploy a DApp's Smart Contracts again right **after**
the tests have been executed. So why is that?
A quick `git blame` (no blame intended here) shows us that this routine has been introduced in https://github.com/embark-framework/embark/commit/12cbb7bdd.
Notice the second list point of the commit:
> Contracts that had been compiled in the JS tests were being deleted at the end of the JS test run, which caused errors
> with the files not being found. The fix was to reset the contracts config to no longer require the non-test contracts
> to be compiled/deployed.
The above is a little tricky to understand without a little bit of context: Embark runs two types of tests, JavaScript tests
and Solidity tests. It does that as part of a single process, keeping state in memory in-between those two test runs.
With that in mind, the commit mentioned above says the following:
1. Artifacts that Embark generates as part of its compilation process are erased after JS tests are done executing.
This makes a lot of sense as we don't want to leave any side effects undone when tests are finished.
2. The commit ensures that not only the artifacts are removed from disc, but also updates Embark's state in memory
for `contractFiles`. The reason for that is that otherwise, in the second test run for Solidity tests, Embark will
throw errors as it tries to look up the path for the artifacts in memory for all the Smart Contracts that had been
compiled before.
3. Last but not least, there's *another* state that needs a reset and that's the Smart Contract configuration. If Embark
doesn't reset the memory it'll assume in the second run that all of those Smart Contracts left in memory
"have no code associated to it", while in reality, they shouldn't be in memory in the first place.
So it seems that `config()` is called a second time with an empty Smart Contracts configuration just to ensure that memory
is reset and no error messages are shown.
As discussed earlier, this unfortunately also introduced a lot of side effects along the way as Embark tries to reregister
ENS subdomains from Smart Contracts that are marked as undeployed and therefore don't have an address to interpolate.
While there's probably different ways to go about it, the most straight forward fix is to simply not call `config()`
a second time when it's really not needed. If the goal is to just reset the memory, we can take advantage of
Embark's internal `config:contractsConfig:set` event, which is what this commit is doing.
The embark-utils package, since 4.1.0-beta.1, is a transitive DApp dependency;
but it's loaded in the main process (not a child process), so embark-utils' own
deps can't be indirectly supplied from embark itself via NODE_PATH.
When introducing the `embark-api-client` in https://github.com/embark-framework/embark/commit/c1bbdbf34
we didn't properly update Cockpits apiService, resulting in calls to API endpoints
that don't exist.
This commit fixes a bug where calls to Embark's API to subscribe to whisper
channels are in a broken state.
It also updates Cockpit's apiService to no longer pass ignored parameters to embark-api-client.
With the move of Embark's modules into their own packages, the names under
which they are registered in the API service have changed as well.
This caused Cockpit to no longer being able to properly detect whether
ENS is enabled in the current Embark project
This package comes with a `setUpEnv()` function that can be called to
prepare Embark's environment variables for different process contexts.
This allows us to set up Embark's environment within a test runner context
without the requirement of importing `packages/embark/src/lib/core/env.js`,
which happens to set up those necessary environment variables as a side effect.
It's important that a path to `packages/embark`'s root is passed down to `setUpEnv()`
to ensure `EMBARK_PATH` gets the correct value.
E.g. within a package `embark-console`, `setUpEnv()` can be called within its tests
like the following:
```
// packages/embark-console/src/test/console.js
import { joinPath, setUpEnv } from 'embark-utils';
setUpEnv(joinPath(__dirname, '../../../embark'));
```
Here, `__dirname + '../../../embark'` ensures `EMBARK_PATH` points to the root
of `packages/embark`. This might look different in other contexts.
For example calling this function from a test file within `packages/embark`, this would
look like this:
```
// packages/embark/src/test/some_file.js
import { joinPath, setUpEnv } from 'embark-utils';
setUpEnv(joinPath(__dirname, '../../'));
```
As part of 880a3a6946, we've introduced a
regression where the `colors` package isn't imported in the scope anymore where it's
needed. In this case the pipeline package. This resulted in log output messages
such as:
```
undefinedwriting fileundefined
```
Adding the `colors` package as dependency and importing it accordingly
fixes the issue.
Make the adjustment even if the test script is currently disabled. Consistently
use nyc to generate coverage with mocha even if the test script is currently
disabled.
If the `embark` cli is detected to be inside the monorepo, the template
generator yarn-links any template deps that are part of the monorepo into the
dapp. However, `embarkjs-connector-web3` now has a dependency on `embark-core`,
which in turn has a dependency on `embark-i18n`. That causes `npm install` to
fail inside the dapp because npm checks for deps of `embarkjs-connector-web3`
relative to the dapp instead of checking relative to the symlink. `yarn
install` does not suffer from that behavior so use it instead.
The demo's `embark.json` relies on the bootstrap package being within the
demo's own `node_modules`, but in the monorepo yarn was hoisting it up to the
root `node_modules`.
Move the debugger module to the `embark-debugger` package.
Handle case where `debug` is entered as a console command, and there is no transaction to be debugged.
add embark-i18n and mocha deps
Add support for `service api on/off` commands.
Deprecate commands `api start/stop` in favor of `service api on/off`.
`service api on` - Enables the API server serving Cockpit. Shows an error if the API server is already starting or started.
`service api off` - Disables the API server serving Cockpit. Shows an error if the API server is already stopping or stopped.
`api start` - This command has been deprecated in favor of `service api on` and will be removed in future versions.
`api stop` - This command has been deprecated in favor of `service api off` and will be removed in future versions.
`api:start` - This event has been deprecated and will be removed in future versions.
`api:stop` - This event has been deprecated and will be removed in future versions.
Remove support for `service whisper on/off` because Whisper cannot be started without first disabling it in the communications config, then restarting Embark.
Add documentation for the `service [process] on/off` command
This command was only added a few commits back so it is not considered a breaking change.
Refactor the `service [process] on/off` commands so that there is only one command. Previously we had a command registered for each process, ie:
```
service blockchain on/off
service ipfs on/off
…
```
Now, we only have one command with many options:
```
service [process] on/off - Starts/stops the process. Options: blockchain, embark, ipfs, api
```
Whisper is deliberately not included in the available options for the aforementioned reasons.
This enables removing unnecessary `core/fs` dependencies which can be
replaced with `fs-extra`.
This commit also fixes a bug introduced in f868d1216d
where methods from `embark.fs` weren't available.
The reason for that is because we have several *Process instances
that are created through child process communication, specifically
process.send() APIs. Using those APIs, we can only send data structures
however, methods attached on any of those will get lost.
This is the case when sending embark.fs through process.send().
We still need fs in those places though, mostly because they are relying
on embarkPath and dappPath().
These places are now importing those functions from `embark-core`. Other
API such as writeFile or mkdirp() can be accessed through fs-extra
or fs modules.
This pull-request upgrades `ethereumjs-wallet`, which has upgraded the underlying dependency on scrypt.js to 0.3.0, making scrypt an optional dependency and offering a pure JS version as a fallback.
The reasoning behind this is that scrypt is problematic to install in some systems, particularly those that don't have node-gyp setup and we have seen some weird issues when installing with elevated privileges (i.e. `sudo npm install -g scrypt`)
This avoids mistakenly placing process event handlers on the parent process
whenever the `embark-core` package is loaded. Also, don't listen for an
`'exit'` event on `process` and then call `process.exit(0)` since the event "is
emitted after the child process ends" ([docs][docs]), i.e. it's unnecessary to
do so and in any case it's not correct to always exit with code `0`.
[docs]: https://nodejs.org/docs/latest-v10.x/api/child_process.html#child_process_event_exit
Add support for ability to start and stop Swarm via command `service swarm on/off`.
Fix issue with swarm not starting due to missing web3 object.
`service swarm on` - starts a swarm node if not already started. Shows an error if the node is already starting or started.
`service swarm off` - kills the running swarm node as long as Embark has started the swarm process. If the swarm process was started externally, an error is shown.
We lost this when we moved the site into the monorepo because
all `dist` folders were ignored.
Co-Authored-By: PascalPrecht <pascal.precht@googlemail.com>
Michael Bradley, Jr
Pascal Precht
snyk-bot
Ricardo Guilherme Schmidt
emizzle
Andre Medeiros
Jonathan Rainville