Setting case-fold-search to nil results "Incorrect handhake from websocket" error about emacs-websocket

[ahyatt]

My fix is committed at 5be01c6, can you try it out?

[millejoh]

I was equally stymied, but now I see it. The debug buffer for the failed connection looks like the following:

[WS] Sending handshake, key: oLL4uXkeq8PgzWDaoRIMXg==, acceptance: NgPYUldwCscZZE51sXufDSBN6hM=
[WS] Received: HTTP/1.1 101 Switching Protocols
Server: TornadoServer/6.0.3
Date: Wed, 16 Oct 2019 14:16:37 GMT
Upgrade: websocket
Connection: Upgrade
Sec-Websocket-Accept: NgPYUldwCscZZE51sXufDSBN6hM=


[WS] Checking for accept header: Sec-WebSocket-Accept: NgPYUldwCscZZE51sXufDSBN6hM=
[WS] Closing websocket
[WS] Closing websocket
[WS] Sending frame, opcode: close payload: nil
[WS] State change to deleted

websocket-verify-headers checks for Sec-WebSocket-Accept, Tornado is returning Sec-Websocket-Accept. Sigh.

Your code does a better job following the RFC than Tornado/Jupter, but I wonder what the chances are of me getting either of those projects to use the proper spelling?

[ahyatt]

This is news to me. I checked the spec again, and it doesn't mention anything about this, in fact the value of Sec-Websocket-Accept is base64, which is case-sensitive.

However, you are correct that the upgrade should be case-insensitive. I'll fix that.

[ahyatt]

Actually, looking at the code again, the check is already case-insensitive (it is wrapped in a (let ((case-fold-search t)) ...) block. Being non-nil, searches and matches should ignore case.

[ahyatt]

Got it. Thanks for the explanation. That I can fix.

[millejoh]

Works for me - thank you!

[ahyatt]

Cool, I merged this into master and will follow up with a version update. I usually push out to the gnu elpa a few weeks after github just to make sure there's no issues.