Important note: We have to specify "On startup" setting to "Continue where you left off" on both clients.

Closing a profile on machine A shouldn't affect the tabs in the list of open tabs on machine B, should it? Nicolas/Sky - is this just a bug? Or is it WAI that we only show currently open profiles' tabs?

I'm not sure, it seems quite ambiguous to me. At first I was worried it was kind of bad that we might leave tabs open after a browser/profile is closed, especially if their "On startup" setting is "Open the New Tab page" or "Open a specific page of set of pages." Like maybe the user intends to perform a lazy/poor mans version of clear browsing data. And not closing these tabs would be unexpected and betray the user's intention.

But the more I think about this, never committing the tab closures until the profile/chrome is started back up seems like the most useful. Obviously is gives more power to restoring foreign tabs, but it also kind of acts as a recently closed tabs section for remote data (in this small use case slice). If the user really wanted to clear the fact that they've visited these tabs, they should clear their history after they close the tabs, I think.

I added some caution on the CL that preceded this bug, https://codereview.chromium.org/2750163003/ , to check the "On startup" setting, but the more I think about this, I don't think we should check the pref.

Curious to hear what zea thinks as well. This might also make it more possible to get into odd sessions states with re-used window/tab ids, as more sync data is persisted through closing the browser. Hopefully your current sessions code handles it correctly.

Yeah, it's an interesting question. My intuition is that, from a desired behavior perspective, we shouldn't treat closing a window/browser/profile as closing each tab individually in that window/browser/profile.

My question was also: do we know what the *currently intended* behavior is (regardless of what it's actually doing?) Have we ever thought about this question before?

Nicolas - what are your thoughts?

Apologies for the delay, hadn't seen this bug till now. For the record, we in the past chose to treat browser closure as different from window/tab closure, since there can be many reasons to close a browser that are unrelated to closing the tabs on that browser.

So I'd agree that that is a bug.

After talking with zea@, it seems unclear what the correct behavior should be when you close the last window for a profile, but Chrome itself stays open, such as when you have multiple profilers or are on OSX. Especailly for the case where you don't have session restore on.

In general though, the current behavior is inherently racey, are we going to get a commit POST off to he server before Chrome's shutdown completes? And it is very inconstent with how sessions restore works. This makes me think, as well, this is something we should change.

Ah, that all makes sense. Should we just say that when Chrome is shutdown, we treat it differently than closing a window or tab? In the specific cases you mentioned (last window of a profile or closing last window on OSX), I think we'd just treat it like a regular window close (i.e. close out the tabs on the server as well).

This bug slipped through the filters of the new Chrome Sync team, sorry for the delay.

Mikel, could you PTAL as this is closely related to SESSIONS you've worked on. Did the USS transition make the bug easier to fix?