500239 - Unable to log in to view knowledge base articles

Status: VERIFIED FIXED

(support.mozilla.org :: General, defect)

Description

[krupa raj[:krupa]]

steps to reproduce:
Note:tester is not logged in
1.Try accessing the link https://support-stage.mozilla.org/en-US/kb/hoova?bl=n
2.Log in page loads
3.provide valid credentials and click "login" button

observed behvaior:
login button is dead-unable to login

additional info:
-behavior can be reproduced while trying to access any KB article
-Normal log in works fine.

screencast:http://screencast.com/t/IgPLUMDk

Comment 1

[E. Cooper [:ecooper]]

webroot/silent_session_start_ajax.php is broken. Since we turned up error reporting on stage, this is probably why were are seeing it now.

I fixed the initial problem, but I now have to work through it because new problems with it are turning up.

Comment 2

[E. Cooper [:ecooper]]

webroot/silent_session_start_ajax.php has always been broken.

http://viewvc.svn.mozilla.org/vc/projects/sumo/trunk/webroot/silent_session_start_ajax.php?view=log

See any revision, line ~7.

Comment 3

[Laura Thomson :laura]

If we turn error reporting back down, does the problem go away?  Because I'm ok with that for now.

Comment 4

[E. Cooper [:ecooper]]

This works on staging again by turning off error reporting via Admin->General.

Looking at SVN (sumo and tiki), I see that this is a SUMO-custom thing. It seems to have always been broken. Fixing the global variable problem mentioned in comment 2 only leads to more errors.

Comment 5

[Laura Thomson :laura]

OK, we'll fix this soon but we don't have to fix it today (phew).

Comment 6

[E. Cooper [:ecooper]]

This needs to be fixed for p10n, since it seems to be a SUMO only thing that has always been broken.

Comment 7

[Paul Craciunoiu [:paulc]]

Seems the behavior is not happening anymore. Fixed?

Comment 8

[Stephen Donner [:stephend] Not actively reading bugmail]

(In reply to comment #7)
> Seems the behavior is not happening anymore. Fixed?

Not happening because of comment 4, right?  (Turning off error reporting.)

Comment 9

[Paul Craciunoiu [:paulc]]

Indeed. My bad.

Comment 10

[Paul Craciunoiu [:paulc]]

https://support-stage.mozilla.org/en-US/kb/this.article.title.has.a.lot.of.periods?bl=n

I don't think it's just webroot/silent_session_start_ajax.php that causes the issue.

Comment 11

[Paul Craciunoiu [:paulc]]

Hey, you think you could clear the cache too?

Comment 12

[Paul Craciunoiu [:paulc]]

(In reply to comment #11)
> Hey, you think you could clear the cache too?
Oops. Wrong bug, sorry :)

Comment 13

[E. Cooper [:ecooper]]

(In reply to comment #10)
> https://support-stage.mozilla.org/en-US/kb/this.article.title.has.a.lot.of.periods?bl=n
> 
> I don't think it's just webroot/silent_session_start_ajax.php that causes the
> issue.

Separate bug. This is a javascript error. The other is a PHP error received via
AJAX. I'll post STR momentarily.

Comment 14

[E. Cooper [:ecooper]]

Steps to reproduce for comment 2 (these need to be followed to view the PHP error):

1) Go to the error reporting section in Admin->General and turn on all security reporting. 'PHP error reporting' level should be at 'Report all PHP errors'. The 'visible to admin only' checkbox should be *unchecked*. Everything else should be on or checked.

2) Log out

3) Visit any staging page or try to edit any page while not logged in

Observed: PHP Error: Undefined variable: prefs
https://support-stage.mozilla.org/silent_session_start_ajax.php?client=all&stub=all
Line 561

Expected: No error. :(

Comment 15

[Vishal Kamdar [:vish_moz]]

Another page where you cannot login:
https://support-stage.mozilla.org/tiki-edit_translation.php?locale=en-US&page=Browsing+basics

Comment 16

[Paul Craciunoiu [:paulc]]

Found a revision that may help:
http://viewvc.svn.mozilla.org/vc?view=revision&revision=10418

Comment 17

[Paul Craciunoiu [:paulc]]

Attachment: patch, v1

Apparently it's not the ajax session start. Gotta love those one line changes :)

Comment 18

[Paul Craciunoiu [:paulc]]

I don't think it's a regression.

Comment 19

[Paul Craciunoiu [:paulc]]

r50317 / r50318

Comment 20

[Stephen Donner [:stephend] Not actively reading bugmail]

This is looking good to me on https://support-stage.mozilla.org/tiki-edit_translation.php?locale=en-US&page=Browsing+basics; I'll let Krupa mark it when she gets a chance.

Comment 21

[krupa raj[:krupa]]

I think there is a regression:

STR:

1.Go to https://support-stage.mozilla.org/en-US/kb/Upgrading+to+Firefox+3%C2%B75
2.Click log in
3.Log in with valid credentials.

Observed behavior:
After successful log in,status still shows "login".However clicking on the "log in" link will inform the user that they are are already logged in.

See the screencast for more details.
screencast: http://screencast.com/t/3rPE1lfwknM

Also see bug 506781

Comment 22

[Tom Ellins [:TMZ]]

Works for me. Try ctrl-F5 to refresh the page. I think you might be hitting the cache.

Comment 23

[Stephen Donner [:stephend] Not actively reading bugmail]

That you could log in means this bug was fixed -- the other bug is separate.

Comment 24

[Stephen Donner [:stephend] Not actively reading bugmail]

Can still reproduce this with an edge case:

http://screencast.com/t/rfGaEalD

Comment 25

[Paul Craciunoiu [:paulc]]

(In reply to comment #24)
> Can still reproduce this with an edge case:
> 
> http://screencast.com/t/rfGaEalD

This works for me locally. Not sure what's going on here.

Comment 26

[Stephen Donner [:stephend] Not actively reading bugmail]

Paul, if you're in tomorrow, I'll do my best to show it to you :-)

Comment 27

[Paul Craciunoiu [:paulc]]

Stephen showed me this and we agreed that it's edge-case-y enough to leave for later.

The problem really is that the ajax way of enabling that "log in" input is not quite foolproof. Filed bug 514206 to find a way of improving this.

Comment 28

[Stephen Donner [:stephend] Not actively reading bugmail]

Verified FIXED; edge case spun off into bug 514206 (in a way; once fixed there, it shouldn't happen, and I tried to reproduce my own edge case and couldn't -- I'm sure it's still there, though).