replayio / playwright Goto Github PK

Playwright pays attention to this env var during installation and doesn't download the browsers. We should do the same thing for replay specific browsers.

https://github.com/microsoft/playwright/blob/75dfc15e6208f26f65ae74c247cf2d0d94820af5/browser_patches/firefox/patches/bootstrap.diff

It looks like installing this package doesn't work in macOS 12.0 Monterey:

$ npm install

> @recordreplay/[email protected] install
> node install.js

Installing replay browsers...
Done.
/Users/dan/devel/playwright/lib/install/installer.js:130
            throw new Error(`Failed to download ${browserName}, caused by\n${e.stack}`);
                  ^

Error: Failed to download chromium, caused by
Error: ERROR: Playwright does not support chromium on mac12-arm64
    at Object.assert (/Users/dan/devel/playwright/lib/utils/utils.js:86:15)
    at Registry.downloadURL (/Users/dan/devel/playwright/lib/utils/registry.js:307:17)
    at Object.downloadBrowserWithProgressBar (/Users/dan/devel/playwright/lib/install/browserFetcher.js:84:26)
    at async validateCache (/Users/dan/devel/playwright/lib/install/installer.js:129:9)
    at async installBrowsersWithProgressBar (/Users/dan/devel/playwright/lib/install/installer.js:80:9)
    at /Users/dan/devel/playwright/lib/install/installer.js:130:19
    at async validateCache (/Users/dan/devel/playwright/lib/install/installer.js:129:9)
    at async installBrowsersWithProgressBar (/Users/dan/devel/playwright/lib/install/installer.js:80:9)

I've successfully integrated Replay with Playwright using this and the recording-cli, but I couldn't find any information on this feature.

I know that Playwright is not a test runner, but it is often used by Jest, or using the official Playwright test runner Playwright Test

It would be great to only keep the recordings / only upload recordings for tests that have failed. This would reduce greatly the quantity of uploads and be easier to sort through them.

Is it currently possible to do something similar to this, even if it requires some glueing?

We updated our playwright fork to 1.18.1 in order to be compatible with @playwright/test but we're hitting a dependency issue.

We can launch the test harness via npx @recordreplay/playwright to override the browser loading to point to the Replay browsers. But @playwright/test requires playwright-core via a child process so when the test launches the browser, it uses the playwright-core logic instead of ours and fails to launch the Replay browser.

I tried to hijack (via Module.prototype.require) the loader logic which worked for part of the flow but fails when the child process loads the actual test file. That process uses the native require (undoctored by my hack) and then launches the default browsers.

It looks like the current implementation requires using playwright.browser.launch() to execute the tests. This requires defining the test run orchestration either within the test file or with a "helper" file and executing the tests by running that file with Node.

The probably more common use case for Playwright is to have a Playwright config file, the test files in the tests directory, using the test syntax to organize test blocks within the file, and then running tests using npx playwright test.

https://playwright.dev/docs/intro#first-test

We should support this implementation if possible.