Bug 32055 – Add test for cross-origin WebSocket usage

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKit Misc.

[Fumitoshi Ukai]

Add many new tests for cross-origin WebSocket usage to check that it functions correctly.

[Alexey Proskuryakov]

Some existing tests (e.g. frame-length-skip) currently use ws://localhost:8880 - these need to be evaluated for correctness, too.

[Fumitoshi Ukai]

Created attachment 44287 [details]
Add tests for cross-origin WebSocket usage

[WebKit Review Bot]

style-queue ran check-webkit-style on attachment 44287 [details] without any errors.

[Alexey Proskuryakov]

Both ws://localhost:8880 and ws://127.0.0.1:8880 are cross-origin if the test itself is served from http://127.0.0.1:8000. So, changing one for another doesn't seem to make much sense.

I got somewhat confused about how origin-based authentication works for WebSocket - unlike XMLHttpRequest, it doesn't have a "only same origin" legacy, so it doesn't matter which server URL to use.

[Alexey Proskuryakov]

And now I just realized that the tests are loaded from ws://127.0.0.1:8880, sorry for the noise.

Maybe we could change the port for websocket tests to not look that similar to port 8000?

[Alexey Proskuryakov]

Comment on attachment 44287 [details]
Add tests for cross-origin WebSocket usage

Looks good, r=me.

One general note: it would be nice if parts of _wsh.py scripts' functionality could be shared between tests. We could use URL queries to pass parameters to them.

Of course, combining them all into one script should not be a goal, but reducing the number of files necessary for an average test would be good.

[Fumitoshi Ukai]

Committed r51743: <http://trac.webkit.org/changeset/51743>

[Fumitoshi Ukai]

(In reply to comment #6)
> (From update of attachment 44287 [details])
> Looks good, r=me.

Thanks for review!

> 
> One general note: it would be nice if parts of _wsh.py scripts' functionality
> could be shared between tests. We could use URL queries to pass parameters to
> them.
> 
> Of course, combining them all into one script should not be a goal, but
> reducing the number of files necessary for an average test would be good.

Agree.
I'll create another patch.