{
"compilerOptions": {
"noImplicitAny": true,
"noImplicitUseStrict": true,
"noEmitOnError": false,
"module": "amd",
"moduleResolution": "classic",
"target": "es5",
"sourceMap": true,
"experimentalDecorators": true,
"types": ["intern"],
"baseUrl": ".",
"paths": {
"abc/*": ["src/abc/*"],
"def/*": ["src/def/*"]
}
},
"exclude": [
"node_modules"
],
"include": [
"src/**/*",
"tests/unit/**/*"
],
"files": [
"./node_modules/dojo-typings/dojo/1.11/index.d.ts",
"./node_modules/dojo-typings/dojo/1.11/modules.d.ts",
"./node_modules/dojo-typings/dijit/1.11/index.d.ts",
"./node_modules/dojo-typings/dijit/1.11/modules.d.ts",
"./node_modules/dojo-typings/dojox/1.11/index.d.ts",
"./node_modules/dojo-typings/dojox/1.11/modules.d.ts",
"./node_modules/dojo-typings/dojox/1.11/html/entities.d.ts",
"./node_modules/dojo-typings/custom/dgrid/1.1/dgrid.d.ts",
"./node_modules/dojo-typings/custom/dstore/1.1/dstore.d.ts"
]
}
You can see tsconfig above.
I do not have separate files for src and tests, there is only one file at the base.
ah, btw it worked with moduleResolution node.
Not sure why it did not work earlier.
I have to check with my lead if this module resolution would work and smoke test the app.
You may be able to use paths, like
{
"compilerOptions": {
"paths": {
"@theintern/common": ["node_modules/@theintern/common/index"],
"source-map": ["node_modules/source-map/source-map"]
}
}However, you may still run into issues because modules published through npm are going to assume node-based module resolution.
Thank you very much for your help.
Funny, someone else just pinged me because GHA recently updated to Chrome 83 and their tests started failing. :grinning: Intern's policy is to track the current stable versions of webdrivers by default. If you need to pin to a specific driver version, you can do that using tunnelOptions.drivers:
{
"tunnelOptions": {
"drivers": [
{ "name": "chrome", "version": "81..." }
]
}
}We can actually make that a bit easier with GHA, because (at least according to their release notes) they publish something to a CHROMEWEBDRIVER environment variable in the container. I haven't checked to see what that contains, though (path to an exe, version number, ...).
This version of ChromeDriver only supports Chrome version 83
It looks like you're using a Windows environment?
Yup, the Windows 2019 environment, which as of seven days ago is using chrome 81 (at least according to this https://github.com/actions/virtual-environments/blame/master/images/win/Windows2019-Readme.md#L573)
Hmm...the release notes for the most recent release (3 days ago) says it's on Chrome 83. (https://github.com/actions/virtual-environments/releases/tag/win19%2F20200524.1)
In any case. the solution is still to pin the chromedriver version to something compatible with the target environment.
It would be useful to be able to easily make use of the environment variable, though. The Windows release notes point out that it points to the driver exe. so that's useful to know. I'll create an issue for it...
Because I can pin to v81 to get tests passing now, but in a few days when the environment updates to v83 I'm gonna have the same issue.
I guess I could, in CI, interpolate the chrome version into the config?
Being able to use environment variables in the config would handle the GHA case (and would be generally useful). You can already use an
INTERN_ARGS environment variable to pass command line options, but it's not very ergonomic for something like this.
Unfortunately, there isn't a way to configure a path to a local webdriver executable, so the best solution for the moment is to pin the version in the config. Depending on the format of their CHROMEWEBDRIVER variable, you might be able to extract the local version from that and use it to set the webdriver version.
So I have unit tests running in the browser. How do I get those to run in a headless browser on my CI server?
@stdavis In a CI system, you'd run Intern from the command line, and use the Selenium tunnel (assuming the browser is on your CI system) to load and manage the browser. In your Intern config, you'd setup an environment for headless chrome (see https://theintern.io/docs.html#Intern/4/docs/docs%2Fhow_to.md/run-tests-with-headless-chrome). Note that your CI system would need Java (headless Java is fine) to run Selenium. If Java isn't an option, you can also interact with chromedriver directly, but it's a bit more work (better support for direct interaction with web drivers is in the works).
1 reply
Is Firefox on GitHub Actions really flaky? I feel like every third run fails with
Timeout: [POST http://localhost:4444/wd/hub/session / {"desiredCapabilities":{"name":"intern","idle-timeout":60,"browserName":"firefox","moz:firefoxOptions":{"args":["-headless"],"browser":"firefox"}}] connection refused
Since upgrading from Intern 3 to Intern 4 we have a number of files that have (much) lower code coverage.
I've been able to chase this down to test cases that use a (dojo) context require with a map to reroute dependencies to mock files. After the test case is finished, we clean up the loaded module (using
It seems that reloading a module with Intern 4 resets the coverage built up so far, whereas with Intern 3 this wasn't the case.
Any thoughts on how to work around this? Do we even consider this a bug?
I've been able to chase this down to test cases that use a (dojo) context require with a map to reroute dependencies to mock files. After the test case is finished, we clean up the loaded module (using
require.undef) to force-load it (without mocked dependencies) the next time it is needed.It seems that reloading a module with Intern 4 resets the coverage built up so far, whereas with Intern 3 this wasn't the case.
Any thoughts on how to work around this? Do we even consider this a bug?