Fix match-ports to handle a null input port

[antosart]

---

Preview | Diff

[antosart]

I have the impression that the match-ports algorithm, as currently written, does not match an empty port with an empty port, while I believe it should.

[annevk]

Another problem here is that default port is an integer, not a string.

[annevk]

Also, the caller can pass in null which isn't the empty string, that's another problem.

[antosart]

I am confused. Looking at where the algorithm is called, a url's port should be either null or an integer. Interestingly, however, a few lines later, although with the same wording, it seems that a url's port (note the different link) is a possibly empty string.

So is a url's port a nullable integer or a string?

[annevk]

It links the wrong concepts. For host too. That should use [=url/port=] I think or <a for=url>port</a>.

[antosart]

Ok, so the inputs of the algorithm are actually a string or null (port A), an integer or null (port B) and a (possibly empty) string (scheme B).

[antosart]

It is however confusing that one of the ports is a string and the other an integer.

[annevk]

In part this stems from the CSP parser not being defined in detail I suspect.

[antosart]

Coming back to this. Parts of the issue is fixed, but the "port-part matches" algorithm is still expecting a non-null possibly-empty string while being called with null or a non-empty string. I've updated the PR to fix that.