Implement a "/shutdown" endpoint

[shs96c]

With reference to SeleniumHQ/selenium#3978. The ChromeDriver and IEDriverServer both understand a /shutdown endpoint that can be used to quit the service. All the official language bindings attempt to call this URL before killing the service process. It might be nice to also implement this in geckodriver.

[andreastt]

There is no /shutdown endpoint in the specification, and we have /session/{session id}/quit to end the session. I would consider it the responsibility of the parent process to stop geckodriver when it is no longer needed.

[shs96c]

/shutdown kills the entire service rather than a single session. It's not in the spec, and the parent processes do kill the geckodriver process.

This endpoint is just how all the language bindings notify the driver service to quit, allowing a controlled shutdown rather than a potential kill -9 That's why I said it might be nice to also implement this is geckodriver :)

[andreastt]

Ideally a parent process should ask nicely with SIGTERM (15) before using SIGKILL (9). This would be functionally equivalent to having an HTTP endpoint for it.

(Practically we don’t have any special shutdown handling in geckodriver yet because signal handling with Rust is not great yet.)

[shs96c]

Sure. That'd be the nice way to do it. None of the language bindings do that, but it'd also be nice (the java ones might…). Feel free to open a reciprocal selenium issue :)

…

Sent from my iPhone

On 7 Jun 2017, at 12:55, Andreas Tolfsen ***@***.***> wrote: Ideally a parent process should ask nicely with SIGTERM (15) before using SIGKILL (9). This would be functionally equivalent to having an HTTP endpoint for it. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

[jimevans]

@andreastt Given that Windows doesn't have SIGTERM and SIGKILL, what's the suggestion for that platform?

[andreastt]

As I said, just terminating geckodriver is absolutely fine. There’s no state in geckodriver that requires cleanup.

[jgraham]

If there is a requirement to have an endpoint here it should be in the spec. But an endpoint for this seems like bad design since you probably don't want random clients to shut down the entire service. Instead that can be in control of whoever owns the process using standard os methods for pprocess control. Closing this, but feel free to reopen if there's anything extra we should do here.

[lock]

This issue has been automatically locked since there has not been any recent activity after it was closed. If you have run into an issue you think is related, please open a new issue.