682844 - Crash during startup results in about:blank saved for each tab, history lost.

Status: RESOLVED INCOMPLETE

(Firefox :: Session Restore, defect)

Description

[Hubert Figuiere [:hub]]

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:7.0) Gecko/20100101 Firefox/7.0
Build ID: 20110824172139

Steps to reproduce:

Firefox crashed several time.
At first it crashed selecting a tab.
Reported, restarted.
Firefox applied the update that was pending.
Firefox retored the session automatically, but when I clicked to close the "what's new" tab, Firefox crashed again.
Reported, restarted.
Firefox show the session crash restore session view, and in the other tab the "what's new" page.
I click restore. Firefox crashed again.
Reported, restarted.
Firefox just crash.
Reported, quit.
I re-open Firefox manually, the session crash restore session view only show "about:blank" twice in the list of tabs to restore.


Actual results:

My session is lost forever.


Expected results:

It should never have lost my session.

Comment 1

[u279076]

Please confirm that the session is/was not restorable via History > Restore Previous Sessions. Also, any chance you can provide the crash reports which caused this issue.

Comment 2

[Hubert Figuiere [:hub]]

Yes the menu was grayed out. Impossible to do History > Restore Previous Sessions

As for the crashes they have been submitted the usual way by Crash Reporter. I don't know how can track them down, but within the hour before I reported that bug, from the same IP address, and using that same email address.

Comment 3

[u279076]

In the location bar type "about:crashes" and press ENTER. You can then copy and paste the links to your most recent reports.

Comment 4

[Hubert Figuiere [:hub]]

Oh nifty ! I didn't know about that one.

Here we go:

http://crash-stats.mozilla.com/report/index/bp-b9cf1e72-f846-4f1d-b399-d8e7b2110829
http://crash-stats.mozilla.com/report/index/bp-a4312626-3419-4dea-b891-a193d2110829
http://crash-stats.mozilla.com/report/index/bp-2bddad9b-eb7c-4a91-9966-666092110829
http://crash-stats.mozilla.com/report/index/bp-ca55efbe-e02c-4641-9a61-8e46c2110829
http://crash-stats.mozilla.com/report/index/bp-f4f98632-16e5-4aba-8e23-149002110829

all withing a 4 minutes window.

Comment 5

[u279076]

These seem to be Graphics related crashes. 
These crashes are probably a separate issue which should be categorized in a separate bug, however, I want to eliminate the possibility of these crashes being a direct cause of this bug.

What URLs were open in tabs?
What add-ons do you have installed?
Are these crashes easily reproducible?
Can you reproduce this on a new profile without any add-ons installed?

Comment 6

[Hubert Figuiere [:hub]]

I didn't think the crashes were related, but think that they caused at the end to lose the session. And at one point the whole window server got killed too....

As for the URL open in the tabs, I forgot. The list was long.

Hasn't happened since.

Comment 7

[u279076]

If this hasn't happened since and cannot be reproduced it will be hard to nail down a fix for this. I'm inclined to resolve this bug INCOMPLETE until we can get more info from you about the specifics causing this to happen.

Comment 8

[Paul O'Shannessy [:zpao] (not bugmail, email directly)]

The crash is bug 682462. This also isn't the only time I've seen somebody say that multiple consecutive crashes at startup caused about:blank, so I'd like to keep this open (perhaps with a note to resolve incomplete later if there's no movement)

Comment 9

[u279076]

(In reply to Paul O'Shannessy [:zpao] from comment #8)
> The crash is bug 682462. This also isn't the only time I've seen somebody
> say that multiple consecutive crashes at startup caused about:blank, so I'd
> like to keep this open (perhaps with a note to resolve incomplete later if
> there's no movement)

How long is long enough do you think? a couple of weeks? months? I can add the CLOSEME whiteboard tag given a date.

Comment 10

[Paul Booker]

"resolve this bug INCOMPLETE until we can get more info from you about the specifics causing this to happen."

+1

Should any unconfirmed bugs that are not likely or impossible to reproduce be marked as INCOMPLETE?

Comment 11

[u279076]

(In reply to Paul Booker from comment #10)
> Should any unconfirmed bugs that are not likely or impossible to reproduce
> be marked as INCOMPLETE?

In general, yes. INCOMPLETE means that we will no longer try to blindly troubleshoot the issue until someone can provide information leading to its confirmation. If there has been no traction or conversation in an unconfirmed bug for a long time we will typically post a final ask for feedback giving the reporter 2 weeks to respond. We track this using the CLOSEME-YYYYMMDD whiteboard tag.

Comment 12

[Paul Booker]

Thanks Anthony, updated my notes.

Best, Paul