The test failures are due to an intermittent parsing error from youtube-dl which I've been seeing fairly frequently for a while. It has been raised multiple times on youtube-dl's issue tracker, and was claimed to have been fixed at least once (obviously not), as far as I know. I never got time to look into what exactly was wrong.
Anyway, the gist of this comment is that there's nothing we can do about at the moment, and it shouldn't block the update.
P.S. I guess it stems from some A/B testing on YouTube's part.
Okay yeah it has an HTTP(S) mode, but since it appears to be used to bypass the GFW here, I'm skeptical that that mode is good enough here. The GFW is pretty good at deep packet analysis. Maybe only the SOCKS mode can sneak through the GFW? At least the last time I had to circumvent the GFW, I used Shadowsocks which is a SOCKS5 proxy. I'm just guessing here. Curl should be able to tell.
By the way, I think we only support HTTP(S) proxies, not SOCKS4/5 proxies. Is XX-net a SOCKS5 proxy? I don't know much about proxies and I can't tell immediately from a Google search.
You sure your proxy actually works with HTTPS connections? What about
```
$ curl -x localhost:8087 https://www.google.com
```
And your Python version? I assume 3.5.2?