Figure out how to pass a Node script to BROWSER

We added support for BROWSER env variable that is interpreted by OS-specific command. On OS X, it runs open http://url.com -a <BROWSER>.

However I’m not sure how to make it execute a Node script, and if it’s even possible. That would be the best for allowing use cases like #1096 and https://github.com/facebookincubator/create-react-app/issues/1450. It would be awesome to investigate how to get it working with a Node script (my naïve attempt at pointing it to a shell script with node env didn’t seem to work), and maybe implement some custom handling for JS extension if we don’t find a solution.

[sunknudsen]

Found a way to achieve this using the script option on MacOS.

The following works if a development profile has been created.

https://support.mozilla.org/en-US/kb/profile-manager-create-and-remove-firefox-profiles#w_starting-the-profile-manager

Set BROWSER=open.js in .env.

// Content of open.js
const { spawn } = require('child_process');

spawn(
  '/Applications/Firefox.app/Contents/MacOS/firefox-bin',
  ['-p', 'development', process.argv[2]],
  { detached: true }
);
process.exit(0);

Then run npm start as usual.

[GAumala]

I think that the best choice is to spawn a child process to execute the node script if the BROWSER string ends in “.js”. It is the easiest way to support al operating systems.

react-dev-utils uses opn to open the browser. This module uses the command open on OS X, start on Windows and xdg-open on other platforms. I only use Linux, so I only know about xdg-open, but I believe the other tools are designed in a similar way.

xdg-open is meant to be used in desktop sessions, like when the user double clicks a file. You click a photo, your photo viewer is opened. You click a video, your media player is opened. In settings, the user can choose what programs to use for each type of file. For security reasons, it is strongly discouraged to execute scripts when double clicked, opening the file with a text editor is a much better choice. This is what probably every Linux desktop environment does by default, since users could be tricked into double clicking malicious scripts.

Handling the “.js” extension is much easier and secure.