43828 - frames page does not properly expire logged-in content

Status: VERIFIED FIXED

(Core :: Networking: Cache, defect, P3)

Description

[Joe Slag]

Basic problem: a page on my company intranet consists of a frameset, and the
contents of the main frame are dynamically generated based on the absence /
presence of a cookie.  If the cookie is absent, a login page is displayed; if
the cookie is present, a page including a logout button is displayed.  Logging
in/ logging out ends by redirecting the browser to the same frameset.

Mozilla always used to handle this redirection properly (up through the build on
the 21st).  the 20000624 build, however, keeps the individual frames cached
after responding to the redirect generated by logging in / logging out.  So,
after selecting "logout" and removing the cookie, the browser still displays the
dynamically generated frame as if the user were still logged in.  The proper
content for that frame can be seen by right-clicking and selecting "reload frame".

Comment 1

[Asa Dotzler [:asa]]

updating component and setting default owner

Comment 2

[Joe Slag]

Same behavior in 2000062610 build

Comment 3

[Eric Pollmann]

Reporter: please attach a test case or URL at which this bug is reproducible.    
Thanks!

Comment 4

[David Krause]

test case: http://endymion.com/products/mailman/demo/mmstdo.cgi

Comment 5

[fig\tree]

1. visit URL, login as showandtell, password: blah
2. logout, login as showandtell2, password: blah
Mozilla  displays content from showandtell account, not showandtell2.

This bug is Linux-only (070608), the bug does not appear on Win32 070608.

bug 40449 is the XP bug for "Cache does not store latest value of pages", I
suggest we come back to this bug after bug 40449 has been fixed.

Bug 44763 makes this bug hard to work around.

Functional Workarounds:
set cache compare mode to "Every time I view the page"
or
Set memory cache and disk cache sizes to 0

(Watch out for bug 44766: debug option Enable Mem Cache does not work)

updating component + adding dependency

Comment 6

[Eric Pollmann]

Giving this to the kind folks over in network land - let me know if I can help!

Comment 7

[Chris Book]

This sounds similar to a problem I am having:
I have a messageboard that is dynamically modified. Normally it works fine.
When another user posts and I reload, I see the message.
If the messageboard is in a frame however, I am unable to view the new message.
It always loads the cached version.

Comment 8

[Darin Fisher]

The test case seems broken... requires that I specify a server.  What is that
all about?  Can someone please update the test case.  Thanks!

Comment 9

[Darin Fisher]

The URL specified in the summary section of this bug report also doesn't
accept the showandtell:blah login.

Comment 10

[basic]

could the issue 1 of Dave's post to bug 81564 be related to this bug? If it is
you have a testcase.

Comment 11

[basic]

The testcase is either:
http://216.246.19.83/products/mailman/demo/mmstdol.cgi
or:
http://endymion.com/products/mailman/demo/mmstdol.cgi

to get it to work, I've created 2 test accounts in yahoo.com

mozilla_test2000 and mozilla_test2001

their password is password

use the above demo with the 2 test accounts with pop.mail.yahoo.com as the server.

I've not been able to reproduce the bug as I'm not on linux. Could someone else
tests this? I've a feeling it has been fixed as the cache system has changed
since this was reported.

Comment 12

[Paul McGarry]

I can't seem to get those accounts to work, but my ISP runs the same software at
http://www.ihug.com.au/cgi-bin/mmstdol.cgi
and I can check my account and log out etc etc without seeing any such caching
problems.
Worksforme Linux 2001070521.

Comment 13

[Paul McGarry]

Got basics urls working after talking to him in #mozillazine.
After following the process given by fig/tree adjusted for basics accounts I get
the expected results. Ie when I log in the first time as mozilla_test2000 I get
emails for mozilla_test2000 and when I log out and then log in as
mozilla_test2001 I see emails for mozilla_test2001.

WORKSFORME!

Comment 14

[basic]

marking as fixed, should have done this long time ago as cache system has changed

Comment 15

[Chris Petersen]

Marking verified per last comments