https-agent makes http PUPPETEER_DOWNLOAD_HOST download fail

[tenuki]

Steps to reproduce

Environment

• Puppeteer version: 4bcdfc9 (master as of June/10/19)
• Platform / OS version: linux debian
• URLs (if applicable): -
• Node.js version: v12.3.1

Steps to reproduce:

http_proxy=http://localhost:3128 PUPPETEER_DOWNLOAD_HOST=http://storage.googleapis.com node install.js

Comments:

• http_proxy above points to a working proxy.
• I execute install.js from an installed version, as a way to reproduce the issue. But the first issue is npm install puppeteer with the same proxy and host as above.
• YES I put http://storage.googleapis.com on purpose, that is when the issue happens. Actually, the repo I pretend to use will be in a http server and that is why I have the problem. In the example, I used googles' as it works for manual download and is enough to reproduce the problem.

Please include code that reproduces the issue.

1. http_proxy=http://localhost:3128 PUPPETEER_DOWNLOAD_HOST=http://storage.googleapis.com node install.js

or:

1. http_proxy=http://localhost:3128 PUPPETEER_DOWNLOAD_HOST=http://storage.googleapis.com npm install puppeteer

What is the expected result?

The http proxy will be used to download the chromium version from the specified $PUPPETEER_DOWNLOAD_HOST host.

What happens instead?

user@host:~/repos/puppeteer$ http_proxy=http://localhost:3128 PUPPETEER_DOWNLOAD_HOST=http://storage.googleapis.com node install.js
ERROR: Failed to download Chromium r666595! Set "PUPPETEER_SKIP_CHROMIUM_DOWNLOAD" env variable to skip download.
Error: Download failed: server returned code 403. URL: http://storage.googleapis.com/chromium-browser-snapshots/Linux_x64/666595/chrome-linux.zip
    at /home/user/repos/puppeteer/lib/BrowserFetcher.js:232:21
    at ClientRequest.requestCallback (/home/user/repos/puppeteer/lib/BrowserFetcher.js:288:7)
    at Object.onceWrapper (events.js:288:20)
    at ClientRequest.emit (events.js:200:13)
    at HTTPParser.parserOnIncomingClient [as onIncoming] (_http_client.js:573:23)
    at HTTPParser.parserOnHeadersComplete (_http_common.js:116:17)
    at Socket.socketOnData (_http_client.js:460:22)
    at Socket.emit (events.js:200:13)
    at ClientRequest.onsocket (/home/user/node_modules/https-proxy-agent/index.js:181:14)
    at Object.onceWrapper (events.js:288:20)
  -- ASYNC --
    at BrowserFetcher.<anonymous> (/home/user/repos/puppeteer/lib/helper.js:111:15)
    at Object.<anonymous> (/home/user/repos/puppeteer/install.js:64:16)
    at Module._compile (internal/modules/cjs/loader.js:774:30)
    at Object.Module._extensions..js (internal/modules/cjs/loader.js:785:10)
    at Module.load (internal/modules/cjs/loader.js:641:32)
    at Function.Module._load (internal/modules/cjs/loader.js:556:12)
    at Function.Module.runMain (internal/modules/cjs/loader.js:837:10)
    at internal/main/run_main_module.js:17:11
user@host:~/repos/puppeteer$

It should be noted that https-proxy-agent isn't meant to be used when the target download is http-served, see:

$ http_proxy=http://localhost:3128 node test.js http://example.com
using proxy server "http://localhost:3128"
attempting to GET "http://example.com"
_http_client.js:142
    throw new ERR_INVALID_PROTOCOL(protocol, expectedProtocol);
    ^

TypeError [ERR_INVALID_PROTOCOL]: Protocol "http:" not supported. Expected "https:"
    at new ClientRequest (_http_client.js:142:11)
    at Object.request (https.js:305:10)
    at Object.request (/tmp/tmp2/node_modules/agent-base/patch-core.js:25:22)
    at Object.https.get (/tmp/tmp2/node_modules/agent-base/patch-core.js:48:21)
    at Object.<anonymous> (/tmp/tmp2/node_modules/puppeteer/test.js:18:7)
    at Module._compile (internal/modules/cjs/loader.js:774:30)
    at Object.Module._extensions..js (internal/modules/cjs/loader.js:785:10)
    at Module.load (internal/modules/cjs/loader.js:641:32)
    at Function.Module._load (internal/modules/cjs/loader.js:556:12)
    at Function.Module.runMain (internal/modules/cjs/loader.js:837:10)