url: handle windows drive letter in the file state

[watilde]

In the whatwg-url, C| should not satisfy the condition to not copy the base's path. This patch also synchronises WPT url test data to verify the update in upstream.

Refs: whatwg/url#305
Refs: web-platform-tests/wpt#5754

Checklist

• make -j4 test
• tests are included
• commit message follows commit guidelines

Affected core subsystem(s)

url

[watilde]

ci: https://ci.nodejs.org/job/node-test-pull-request/7824/

[rmisev]

Oops! It looks like you copied other tests ... which was already in your tests file. Correct ones are commented # Windows drive letter handling with the 'file:' base URL in the https://github.com/w3c/web-platform-tests/blob/f0b286344dd2613d48643394408f4ab48ef2b33d/url/urltestdata.json#L5426-L5524

[watilde]

Opps sorry! I re-copied them and got an error in make-test for this part(the other cases worked fine):

  {
    "input": "C|",
    "base": "file://host/dir/file",
    "href": "file:///C:",
    "protocol": "file:",
    "username": "",
    "password": "",
    "host": "",
    "hostname": "",
    "port": "",
    "pathname": "/C:",
    "search": "",
    "hash": ""
  },

I will work on it tonight. Thanks for pointing it :) @rmisev

[refack]

CI: https://ci.nodejs.org/job/node-test-commit/9600/
(Interested how it behaves on Windows...)

[rmisev]

The remaining means after ch, so p[0] must not be counted here. Correct to end - (p+1) == 0 or just end - p == 1

There is an example, which explains the term remaining: https://url.spec.whatwg.org/#remaining

[watilde]

I see it works! Thanks for your comment :)

[rmisev]

The same as above: end - (p+1) >= 2 or just end - p >= 3

[addaleax]

The last CI doesn’t look so good… :/

[watilde]

@addaleax Yeah sorry, I was WIP at the time without the wip label. I pushed the fix with the review comments, but will recheck it tonight.

[addaleax]

Okay, new CI then: https://ci.nodejs.org/job/node-test-commit/9608/ :)

[TimothyGu]

Depending on the spec writer's intention (whatwg/url#308) changing end - p == 1 to end - p <= 1 might be more robust to cover ch == kEOL as well, but in any case it shouldn't matter since the case when ch == kEOL is covered in another case already. Also, nice catch on noticing the end - p == 1 in the original code should actually be end - p == 2 👍

[TimothyGu]

Seems like they agree with me.

We can also calculate per iteration an actual remaining variable, like we are already doing for some other functions in the file. This should eliminate any confusion around this variable. The most correct implementation would be something like const size_t remaining = end == pointer ? 0 : (end - pointer - 1);.

[watilde]

Woo that'd be nice! I'm going to update it 😺

it's gonna be like

const size_t remaining = end == pointer ? 0 : (end - pointer - 1);

......

if ((remaining == 0 ||
      !IsWindowsDriveLetter(ch, p[1]) ||
      (remaining >= 2 &&
       p[2] != '/' &&
       p[2] != '\\' &&
       p[2] != '?' &&
       p[2] != '#'))) {

[watilde]

I replaced end - pointer with remaining I added.

[watilde]

ci: https://ci.nodejs.org/job/node-test-pull-request/7851/

[TimothyGu]

LGTM!

[TimothyGu]

Landed in 943dd5f.