Adding PORT to domain name causes WebPositive to do a Google search instead of load the web page

[codewrangler]

Steps to Reproduce

1. Boot hrev48639 of HaikuOS
2. Launch Web Positive
3. Type 'server.thursby.com:8080' into address bar and hit enter
4. Notice that Web+ does a google search instead of recognizing it as a valid URL.

[codewrangler]

Screenshot of issue

[pulkomandy]

You can "force" it to work by adding a protocol: http://server.thursby.com:8080

[codewrangler]

Yes, you can force it...However, you shouldn't have to. Web+ should have better logic to match the behavior of popular browsers like Safari, Chrome, Firefox and IE.

For example.... 'domain.com:8080' is a valid URL. As it's being typed into a 'browser' address bar, if it can't determine the protocol and it looks like a URL, it should try '​http://', since that's MOST likely what the user meant, since it's a Web browser.

[leavengood]

The code I originally wrote for this was very stupid, though it seems to have been improved a lot in subsequent years:

​http://cgit.haiku-os.org/haiku/tree/src/apps/webpositive/BrowserWindow.cpp#n2465

We probably just need to create a new class to handle this and mimic Chrome's logic:

​https://code.google.com/p/chromium/codesearch#chromium/src/components/omnibox/autocomplete_input.cc&l=132

The above is pretty complicated (and maybe more complicated than we need), but it is well commented.

Having some unit tests around it would be good probably.

This would be a great little piece of code for someone to contribute!

I would enjoy doing it but don't have the time.

[leavengood]

Though having said all that, this particular specific bug of having a port number is probably easy enough to fix without going too crazy.

[codewrangler]

Replying to leavengood:

Though having said all that, this particular specific bug of having a port number is probably easy enough to fix without going too crazy.

I would agree...

[pulkomandy]

I mentionned the workaround only so you can access your website until this get fixed. Of course we need a proper fix.

The BUrl class has some rather good URL parsing now (and it has a test suite). Maybe we could make more use of it.

[pulkomandy]

Fixed in hrev48764.