Current cherrypy 18.0.0 fails to open websocket connection with ws4py 0.5.1

[ralhei]

• I'm submitting a ...

• [x ] bug report
• feature request
• question about the decisions made in the repository

• What is the current behavior?
  The browser (firefox) fails to make a connection to the /ws endpoint. Downgrading to cherrypy 17.3.0 or earlier doesn't show this behavior, everything works fine there.

• If the current behavior is a bug, please provide the steps to reproduce and if possible a screenshots and logs of the problem. If you can, show us your code.

• Please tell us about your environment:

• Cheroot version: 6.5.2
• CherryPy version: 18.0.0
• Python version: 3.6.4
• OS: centos7
• Browser: Firefox

• Other information (e.g. detailed explanation, stacktraces, related issues, suggestions how to fix, links for us to have context, e.g. stackoverflow, gitter, etc.)

[webknjaz]

More details needed. Currently there's no reproducer.
cc @Lawouach

[Lawouach]

Hey there,

I will give it a look but the ws4py is considered a project that has not been worked on for a few years now so I'm not surprised considering it was tapping into the internals of CherryPy.

[Lawouach]

Okay, so the issue is that CherryPy response headers seems to now be unicode when ws4py was passing byte objects.

Could someone confirm me this is the case with cheroot and when this change happened? So I can add the right fix in place in ws4py.

[webknjaz]

@Lawouach I bet it's CherryPy switching to Python 3.5+ starting v18.0.0, but might be wrong.

cc @jaraco

[jaraco]

Hmm. If I recall correctly, the change for v18.0.0 was mainly in metadata. Here are the differences.

[jaraco]

I suspect the issue is with d3fa6b8, as text_or_bytes is different from string_types on Python 3.

[webknjaz]

Yeah, that's also my guess here.

[jaraco]

Question is - do we wish in CherryPy to restore support for passing headers as bytes? I think we'd like to get away from supporting both bytes/text in a variety of interfaces. If we do wish to support this use case, we should write a unit test to capture that expectation. But my instinct is we should ask downstream consumers to supply text headers.

[webknjaz]

Oh, wait. I've merged this recently: #1736

[webknjaz]

@ralhei @Lawouach can anyone verify whether it's reproducible on master?

[Lawouach]

Just tried. Couldn't see the same problem with master indeed.

[webknjaz]

So we just need to cut a new patch version then :)

[Lawouach]

I think that would do :p

[jaraco]

Releasing as v18.0.1.