Fixed parse_url(): can not recognize port without scheme

[PandaLIU-1111]

fixed Before

var_dump(parse_url("127.0.0.1:9999"));
var_dump("==========");
var_dump(parse_url("127.0.0.1:9999?"));

output

array(2) {
  ["host"]=>
  string(9) "127.0.0.1"
  ["port"]=>
  int(9999)
}
string(10) "=========="
array(3) {
  ["scheme"]=>
  string(9) "127.0.0.1"
  ["path"]=>
  string(4) "9999"
  ["query"]=>
  string(0) ""
}

# fixed After

var_dump(parse_url("127.0.0.1:9999?"));

output

array(2) {
  ["host"]=>
  string(9) "127.0.0.1"
  ["port"]=>
  int(9999)
 ["query"]=>
  string(0) ""
}

[twose]

https://3v4l.org/RBvOW
It shows when we add the scheme, its behavior back to normal, so I think it is correct to have the same behavior in the non-scheme branch.

[twose]

Maybe we should fix it by this way on line 153:

if ((p == ue || *p == '/'  || *p == '?') && (p - e) < 7) {

[cmb69]

Thank you for the PR! I'm not against fixing this particular issue for partial URLs (although https://3v4l.org/vtrZb should probably fixed as well), but I consider it more important to have an URI parser which strictly conforms to the specification, and which doesn't even attempt to parse partial URIs (that would need to be another function for BC reasons, though).

[PandaLIU-1111]

so you think we need create an URI parser to replace parse_url? @cmb69

[cmb69]

so you think we need create an URI parser to replace parse_url?

Yes, I think we need a proper URI parser, but not to replace parse_url(), but rather in addition to parse_url(). In the long run parse_url() might be deprecated and removed, but we need to keep it for now for backward compatibility reasons.

[PandaLIU-1111]

ok, i prepare to copy parse_url method code and change it for fix scheme , so how we name the method ? parse_uri? @cmb69

[cmb69]

The name of the new function can still be decided later. But I wouldn't reuse the code in ? parse_url()`, because in my opinion, it is broken beyond repair.

[PandaLIU-1111]

@cmb69 hello, How progress this pr?

[cmb69]

Thanks for the reminder! I have merged the PR after renaming and slightly tweaking the test case.

[Berdir]

This caused a regression in Drupal, where we support URL's that look like this: internal:#foo, now internal is recognized as a path and not as a scheme anymore: https://3v4l.org/nFqLN