Closed Bug 602746 Opened 9 years ago Closed 9 years ago

Recent cookie changes made them session only

Categories

(Core :: Networking: Cookies, defect, major)

defect
Not set
major

Tracking

()

RESOLVED FIXED

People

(Reporter: matjk7, Unassigned)

Details

(Keywords: regression)

User-Agent:       Mozilla/5.0 (Windows NT 6.1; rv:2.0b8pre) Gecko/20101007 Firefox/4.0b8pre
Build Identifier: Built from http://hg.mozilla.org/mozilla-central/rev/007411b98478

Regression from either bug 591447, bug 595925, bug 598196 or bug 599799. This is pretty bad and should block beta8.

Reproducible: Always

Steps to Reproduce:
1.Use any build after rev http://hg.mozilla.org/mozilla-central/rev/8835848bbd17
2.Close the browser
3.Go to a site you were logged in.
Actual Results:  
Site asks for login information (corrupt/missing cookies).

Expected Results:  
Cookies should work.
blocking2.0: --- → ?
Keywords: regression
Version: unspecified → Trunk
I really don't understand this. All the persistent unit tests we have pass fine, which means cookies are getting written out and read in just fine.

This could be Windows-only; I'll spin up my builds there and test.

For now, I'm going to back everything out.
Status: UNCONFIRMED → RESOLVED
blocking2.0: ? → ---
Closed: 9 years ago
Resolution: --- → FIXED
Just want to confirm that my cookies were all lost.  I had to delete cookies.sqlite.

Mozilla/5.0 (Windows NT 6.1; WOW64; rv:2.0b8pre) Gecko/20101007 Firefox/4.0b8pre - Build ID: 20101007162259
(In reply to comment #2)
> Just want to confirm that my cookies were all lost.  I had to delete
> cookies.sqlite.
> 
> Mozilla/5.0 (Windows NT 6.1; WOW64; rv:2.0b8pre) Gecko/20101007
> Firefox/4.0b8pre - Build ID: 20101007162259

btw... after deleting cooklies.sqlite, cookies are working fine again.
Interesting. So this is something migration-related, probably Windows-only, and perhaps due to WAL.

Has anyone seen this on non-Windows?
(In reply to comment #2)
> Just want to confirm that my cookies were all lost.  I had to delete
> cookies.sqlite.
Getting a copy of that file would have been useful.  In the future, don't just delete things, but please move them aside.
Can you also provide exact steps of what you did, while they're in your head?

I'm assuming it was the following:

1) Download & install hourly
2) Start browser
3) Cookies worked fine, still logged into all forums etc
4) Close browser
4) Start browser, cookies are gone.

Is this correct?
All cookies lost after restart Minefield on Linux.
http://hg.mozilla.org/mozilla-central/rev/8835848bbd17
chrome://browser/content/aboutDialog.xul Firefox/4.0b8pre ID:20101007140706
We definitely want some tests to cover this case, methinks.
(In reply to comment #8)
> See comment 6?
Yes, I did. I checked the following site.
https://bugzilla.mozilla.org/ and http://forums.mozillazine.org/

And you can check stored cookies by following procedure.
5) Alt > Edit > Preferences > Privacy > select "User custom settings for history" > click "Show Cookies..."
After I installed this fix I restarted Fx4.  I immediately noticed that my cookies were not being 'used'.  I then proceeded to go to sites where I want my cookies saved.  I logged in to a few of them that keep my login info in a cookie.

I looked at my cookies in Options and they were there.  However when I got out of Fx and then back in all cookies were lost. The only way I got cookies to 'stick' was to delete cookies.sqlite and fire up Fx again.  Now my cookies 'stick'.

The cookies file was filled with all my cookies when I looked at the database with a program of mine.  However Fx didn't see them after applying this fix.
(In reply to comment #6)
> Can you also provide exact steps of what you did, while they're in your head?
> 
> I'm assuming it was the following:
> 
> 1) Download & install hourly
> 2) Start browser
> 3) Cookies worked fine, still logged into all forums etc
> 4) Close browser
> 4) Start browser, cookies are gone.
> 
> Is this correct?

Comment 0 was slightly incorrect, as soon as I install a build with the cookie changes they are busted. I still have a broken copy of cookies.sqlite if its needed. I'll see if I can reproduce this with a new copy of cookies.sqlite, and if so get a cookie log.
(In reply to comment #12)
> Comment 0 was slightly incorrect, as soon as I install a build with the cookie
> changes they are busted. I still have a broken copy of cookies.sqlite if its
> needed. I'll see if I can reproduce this with a new copy of cookies.sqlite, and
> if so get a cookie log.

OK, that makes more sense. Thanks!

The broken copy would be useful -- can you email it to me? By "broken" you mean a file created on an older build and then used once with a new build, right? Log probably won't help much.
Why this bug is marked  resolved fixed?
Because the changes got backed out. It's fixed.
You need to log in before you can comment on or make changes to this bug.