468301 - Non-private session is restored after crashing in Private Browsing mode

Status: RESOLVED INVALID

(Firefox :: Session Restore, defect)

Description

[Martijn Wargers (dead)]

To reproduce:
- Go into Private Browsing Mode
- Crash while being in Private Browsing Mode

When restarting, you get to see the session restore dialog, which indicates that you crashed the last time.
I don't think that should be happening, right?

Comment 1

[Nochum Sossonko [:Natch]]

Also related is bug 463188, although shouldn't it restore the pre-private browsing session in the case of a crash?

Comment 2

[(no longer active)]

(In reply to comment #0)
> To reproduce:
> - Go into Private Browsing Mode
> - Crash while being in Private Browsing Mode
> 
> When restarting, you get to see the session restore dialog, which indicates
> that you crashed the last time.
> I don't think that should be happening, right?

I don't understand what you expect to happen.  Does the restored session include anything from the private session?  (it shouldn't)

Or do you mean that we should ignore the crash from within the private browsing mode?  If so, why?

Comment 3

[(no longer active)]

(In reply to comment #1)
> Also related is bug 463188,

Why is that relevant?

> although shouldn't it restore the pre-private
> browsing session in the case of a crash?

Yes, it should.  Are you suggesting that it doesn't?

Comment 4

[Nochum Sossonko [:Natch]]

Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9.2a1pre) Gecko/20081206 Minefield/3.2a1pre

For me it restores the pre-private browsing session, I just thought that should be expected.

Comment 5

[(no longer active)]

(In reply to comment #4)
> Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9.2a1pre) Gecko/20081206
> Minefield/3.2a1pre
> 
> For me it restores the pre-private browsing session, I just thought that should
> be expected.

Yes, it is working as expected.  Thanks for verifying this.

Comment 6

[Martijn Wargers (dead)]

For me, it doesn restore the pre-private browsing session.
Except it gives me the session restore dialog, with the pages of the pre-private browsing session.
That doesn't seem to make sense to me.

It seems to me the crash from within the private browsing session should be entirely ignored, like suggested at the end of comment 2 
Or perhaps on restart, the session restore dialog should come up with the option of restoring the pages you visited from within the private browsing session (but that seems like the wrong thing to do to me).

Comment 7

[Martijn Wargers (dead)]

Steps to reproduce:
- Start Firefox, visit some web pages, wait at least 10 seconds
- Go into Private Browsing Mode, visit some web pages, wait at least 10 seconds
- Visit https://bugzilla.mozilla.org/attachment.cgi?id=351376 (testcase from bug 467914 that crashes Firefox trunk)
- Breakpad client comes up, choose Restart Firefox

After that, the restore dialog of the pre-private browsing session comes up.

Comment 8

[Simon Bünzli]

Martijn: Make sure that browser.sessionstore.max_resumed_crashes is set to 1 and that you don't crash Firefox right before step 1 from comment #7.

Alex: This is the same as bug 463188 except that instead of quitting regularly we've crashed instead. Should this case be treated differently?

Comment 9

[(no longer active)]

I think this is INVALID, since by design we don't store the private session anywhere on disk, and we don't want the browser to have different behavior after being done with the private mode.  Here, restoring the previous (non-private) session is the usual behavior, so I think PB should not break it.

Comment 10

[Martijn Wargers (dead)]

(In reply to comment #8)
> Martijn: Make sure that browser.sessionstore.max_resumed_crashes is set to 1
> and that you don't crash Firefox right before step 1 from comment #7.

Yes, I have that.

(In reply to comment #9)
> I think this is INVALID, since by design we don't store the private session
> anywhere on disk, and we don't want the browser to have different behavior
> after being done with the private mode.  Here, restoring the previous
> (non-private) session is the usual behavior, so I think PB should not break it.

Currently, you have different behavior, because after you've crashed within the private browsing mode, you get the session restore dialog upon restart, while you wouldn't get that if the private browsing mode was normally ended.

Comment 11

[:Gavin Sharp [email: [e-mail]]]

(In reply to comment #10)
> Currently, you have different behavior, because after you've crashed within the
> private browsing mode, you get the session restore dialog upon restart, while
> you wouldn't get that if the private browsing mode was normally ended.

I don't understand why the different behavior is a problem. Something different occurred, after all (you crashed). Why should crashing private browsing mode behave the same way as ending it normally?

Comment 12

[Martijn Wargers (dead)]

Ok, I'll resolve this to invalid then.

Comment 13

[(no longer active)]

Regarding comment 10:

Simon: what causes session store not to show the about:sessionrestore page after a crash in normal mode?  Maybe we should simulate what happens there in the private browsing mode as well.

Comment 14

[Simon Bünzli]

(In reply to comment #13)
Actually, I don't get about:sessionrestore when crashing in PB mode, either - and neither should you (unless you've just crashed Firefox before or set browser.sessionstore.max_resumed_crashes to 0).

Comment 15

[(no longer active)]

Martijn: had you crashed before observing this?  Can you reproduce what you saw in a clean profile?

I guess you can detect the number of recent crashes by opening sessionstore.js in your profile and looking up the value of the recentCrashes attribute.

Comment 16

[Martijn Wargers (dead)]

Yes, I can reproduce in a clean profile.

It says:
recentCrashes":2

So I my steps to reproduce weren't good. You have to crash twice to reproduce the problem (something to do with session restore only showing the session restore dialog on the second crash).

Comment 17

[Simon Bünzli]

(In reply to comment #16)
> You have to crash twice to reproduce the problem

That's the expected behavior (which I tried to tell you in slightly cryptic form in comment #8: as soon as you crash Firefox before your STR, you should get the new Session Restore page), now all is well.