1566243 - Columns 'OpCode', 'MaskBit' and 'FinBit' are showing wrong values

Status: RESOLVED FIXED

(Core :: Networking: WebSockets, defect, P3)

Description

[Christoph Walcher]

Attachment: Screenshot with comments

1. Column 'OpCode' shows wrong opcode
2. Column 'MaskBit' is always true, even if the payload is not masked
3. Column 'FinBit' is always true. The WS engine automatically combines fragmented frames, so this Column can't be false?

Also the ping request of the server is displayed after the response from the browser

Code I used for testing this: https://gist.github.com/wiomoc/00bfdb866e3f5383bc8d302d24a4d43f

Comment 1

[Jan Honza Odvarko [:Honza] (always need-info? me)]

@hengyeow: Can you please test this and see whether it's the platform who's providing the wrong data?
(+ also check the order of WS frames)

Thanks,
Honza

Comment 2

[Hengyeow (:tanhengyeow)]

Attachment: websocket-actor-frameReceived-callback-logs.png

Hi @Honza

I was able to reproduce the scenario mentioned. The attached image shows a series of console logs of frame details (provided by platform) intercepted by the frameReceived callback function in websocket-actor.js.

Comment 3

[Jan Honza Odvarko [:Honza] (always need-info? me)]

@baku, it looks like the platform is giving us wrong WS frame data.
Can you please have a look?

Thanks,
Honza

Comment 4

[Andrea Marchesini [:baku]]

Can you assign this to someone from the DOM team? It's important for devtools, but I'm not able to work on this right now. Thanks.
In case, I can give an overview about how webSocket events for devtools are dispatched.

Comment 5

[Christoph Walcher]

Looks like in https://searchfox.org/mozilla-central/source/netwerk/protocol/websocket/WebSocketFrame.cpp#113 the parameters 'opcode' and 'maskBit' are serialized in a different order then deserialized.

Comment 6

[Christoph Walcher]

Attachment: Bug 1566243 - Fix marshalling of WebSocketFrame

Comment 7

[Hsin-Yi Tsai (she/her) [:hsinyi]]

(In reply to Andrea Marchesini [:baku] from comment #4)

Can you assign this to someone from the DOM team? It's important for devtools, but I'm not able to work on this right now. Thanks.
In case, I can give an overview about how webSocket events for devtools are dispatched.

Taking via email discussion.

Comment 8

[Nhi Nguyen (:nhi)]

Thanks for submitting the patch, Christoph! Assigning the bug to you.

Michal, could you help submit the patch if it's correct? Thanks!

Comment 9

[BugBot [:suhaib / :marco/ :calixte]]

The product::component has been changed since the backlog priority was decided, so we're resetting it.
For more information, please visit auto_nag documentation.

Comment 10

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/e358fa3061e5
Fix marshalling of WebSocketFrame r=michal

Comment 11

[Stefan Hindli [:stefan_hindli]]

https://hg.mozilla.org/mozilla-central/rev/e358fa3061e5