looks good to me, there is some good inspiration over at https://github.com/uriparser/uriparser/blob/master/test/test.cpp
for other test variants.

Is there any reason why const isn't used in this prototype, for URL?:

CURLUcode curl_url(char *URL, CURLURL **urlhandle, unsigned int flags)

This function makes a call to parseurl(), where a const could added as well, and where internally the buffer isn't written nor passed to functions that write into it. ...it's possible I missed something reading the code.

UPDATE: Forgot to say: Thanks for this API, it is very useful!

No reason at all really. It's just an old weakness of mine to skip out on the 'const's... I'll fix. Thanks!

More minor fixes, squashed and rebased.

I think coveralls got some hiccups there when it claims a -3.4% less test coverage, as this patch provides quite a bit of testing already. I'm ignoring that for now, as I hope to add more tests anyway as I go along and fix the remaining parts.

I'm not entirely happy with how curl_url() works just now, as just now we have two ways we can set a URL to the handle. I'm switching from:

Current

  CURLURL *h;
  rc = curl_url("http://example.com", &h, 0); /* create handle *

I suspect that way of storing the handle pointer is going to cause users grief. A bit too unusual.

New

  CURLURL *h = curl_url(); /* create handle */
  curl_url_set(h, CURLUPART_URL, "http://example.com", 0);

It is more in style with how our other APIs work and makes sure there's one single way to set URL in the handle,

Syntax-check a URL

A downside with this approach is that there's no longer any "shortcut" for just syntax-checking a URL without creating a handle, but then again I'm not convinced that's a very important feature.

I've synced the wiki page to match how the current patch implements the API.