[Desktop] Read WebDriver BiDi port from WebDriverBiDiActivePort file within the Firefox profile directory
Categories
(Testing :: geckodriver, task, P2)
Tracking
(Not tracked)
People
(Reporter: whimboo, Unassigned)
References
(Blocks 1 open bug)
Details
Similar to CDP (bug 1706581) and Marionette (bug 1735162) we should do that for BiDi. Hereby I wonder if we should use a separate file or combine it with Marionette to lets say have a single file with the following contents:
BiDi: localhost:9992
Marionette: localhost:2828
If we wanna do that we should do it now before we actually make use of it via bug 1240830.
Otherwise we will just put localhost:9992
in a WebDriver BiDi specific file within the profile.
Reporter | ||
Comment 1•3 years ago
|
||
We discussed today and agreed to have separate files. It will actually be hard to write a single file given that both Marionette and WebDriver BiDi are different components and can be enabled/disabled individually. Maybe there is some cost to create/write to a second file but it's only happening when --remote-debugging-port
has been specified.
Reporter | ||
Updated•3 years ago
|
Reporter | ||
Comment 2•3 years ago
|
||
Given that this feature would be helpful for Selenium integration lets try to get it uplifted to 97 beta if still possible.
Reporter | ||
Updated•3 years ago
|
Reporter | ||
Updated•3 years ago
|
Reporter | ||
Updated•3 years ago
|
Reporter | ||
Comment 3•3 years ago
|
||
This is for Desktop right now. Android will follow-up later on.
Reporter | ||
Comment 4•3 years ago
|
||
I was thinking more about that issue and actually there is nothing to do for us on geckodriver's site. Let me explain why:
geckodriver itself allows consumers to opt-in to WebDriver BiDi by specifying the webSocketUrl
capability. Remote Agent will then create the WebSocket server on the given port, which Marionette includes in the returned webSocketUrl
capability. Given that this is returned to consumers of geckodriver they are already aware of the correct port.
The only problem I could foresee is with Android. Given that the connection is not a plain socket connection and contains the host header (including the port), the adb forward has to use the same port on both sides (host and device). As such using a port of 0
could create a race condition if the port is already taken on the host side. But that's clearly a separate issue.
In case we all agree I would close this bug and file another one for the Android forwarding case.
Reporter | ||
Comment 5•3 years ago
|
||
As discussed in the meeting we will close this bug and I'm going to file a new bug for Android.
Reporter | ||
Updated•3 years ago
|
Description
•