Node should serve files directly to puppeteer

[samreid]

From #90

Having a hardcoded slash after the localTestingURL was causing bugs for me on Windows http-server, and this was not how we used localTestingURL when generating PhET-iO API files. I removed that hard coded slash, and added support for dynamically adding it if it isn't provided from the client's build-local.

Another way to address this would be to have node serve the files to puppeteer instead of relying on the developer localhost. We implemented this for https://github.com/phetsims/phet-io/issues/1664 and the same should be possible here.

[samreid]

Let's try to factor out common code between this and phetsims/chipper@9d74fe9

[samreid]

The usages of localTestingURL in aqua/Gruntfile.js fall more under the CT umbrella, and I can't tell whether it would be appropriate to use http.createServer in those cases. @jonathanolson what do you recommend?

[samreid]

The other part for aqua is in

aqua/js/local/puppeteerQUnit.js

Line 85 in 0cf643e

await page.goto( `${targetURL}&qunitHooks` );

[zepumph]

In phetsims/perennial#175 (comment) I was surprised that I found that when using http.createServer to load a sim in binder, a bug in supplying the MIME type (Content-Type header) broke the sim. Why does this not occur with the Continuous Server? perhaps because I was specifying an svg incorrectly as a plain/text. Do you think this server could benefit from using withServer? Or does it need to exist longer than that function is set up to?

[zepumph]

In regards to #97 (comment), this no longer uses localTestingURL because of its usage change over in #97 (comment) in hook-pre-commit to use withServer, so I think this is really just about ContinuousServer.js

[jonathanolson]

The usages of localTestingURL in aqua/Gruntfile.js fall more under the CT umbrella, and I can't tell whether it would be appropriate to use http.createServer in those cases. @jonathanolson what do you recommend?

Seems possible to serve that way, but would it be a performance hit? Seems nice to factor it out so it's not needed if possible.

[samreid]

I'm not very familiar with it, but it seems that ContinuousServer is supposed to last longer than the withServer is designed for. @zepumph what do you recommend?

[zepumph]

ContinuousServer.startServer begins and runs until the program stops. Sounds like we should leave it as is. I think this issue is outdated and should be closed.

[samreid]

Closing sounds good to me.

[zepumph]

I also want to add that we now have a puppeteer-based client for CT, see #143