Handle Target life time about devtools-window

[harthur]

I think the toolbox should close without prompting if it's in the sidebar or bottom of the browser.

[paulrouget]

Yeah. I don't know yet what to do for the frontend part. But for sure, we need (cancelable?) events to handle the destruction of a target.

[paulrouget]

Maybe a target should be a more complex object that can fire events. (just a thought).

[paulrouget]

Events the target should fire:

• destroy (tab closed)
• hide / show (tab switching)
• unload (before reload or tab close)
• load (once a new root document (tab.linkedBrowser.contentWindow/contentDocument) is available)

Probably more to come when we'll support remote targets.

[joewalker]

I've suggested an updated target API here: #72 perhaps that helps with this question.

[harthur]

Thanks Joe, I'll try this out in our Toolbox code.

[paulrouget]

Update: this is what I think we need:

• will-close (optional and cancelable)
• close
• will-navigate (optional and cancelable)
• navigate

I don't think we need reload and change.

[paulrouget]

I'll take care of that.

[past]

For remote targets, close will have to fire when a tabDetached packet is received, but I don't think we can implement a cancelable will-close, since we don't have that kind of control over the remote process. The web console has a locationChanged packet that is set twice, mimicking will-navigate/navigate, but the debugger only has a single tabNavigated packet that corresponds to navigate.

We have a bug on file to use the web console's behavior with the tabNavigated packet type though. I don't believe locationChanged/tabNavigated is cancelable, either and I doubt such a protocol change would be welcome.

[paulrouget]

"Cancelability" is optional. will-close and will-navigate are optional.

[paulrouget]

Update, for tabs, I'm implementing these events:

close
will-navigate (cancelable)
navigate
hidden
visible

The mandatory events should me:

close
navigate

The target, after firing "close", should destroy itself.

[paulrouget]

fixed by PR #231