40349 - accessing secure page hangs browser

Status: VERIFIED DUPLICATE of bug 41680

(Core Graveyard :: Security: UI, defect, P1)

Description

[sairuh (rarely reading bugmail)]

found this using linux [commercial] bits 2000.05.23.13. i noticed this failing
y'day (might've been sooner). however, linux is still able to pass the security
smoketest of going to https://digitalid.verisign.com/, but i'm confused as to
why it fails with our internal secure server.

here are my steps:

1. in the web browser, type 'calendar-1' (without the quotes ;-) in the URL bar.
2. hit the enter key.
3. in the Security Warning dialog that appears, click OK.
4. watch the throbber and progress meter go on forever...the page remains
blank...
5. after 2min, i impatiently click the Stop button.

result: the throbber and progress meter stop animating. but now the browser is
frozen! unable to select widgets. need to kill the window, then kill the pid.
here's the console output:

Entry at index 0 is calendar-1
->>>>>>>>>>>>>> Write Clipboard to memory
->>>>>>>>>>>>>> Write Clipboard to memory
Document: Done (0.373 secs)
Document http://calendar-1/ loaded successfully
WEBSHELL+ = 5
WEBSHELL+ = 6
WEBSHELL+ = 7
commonDialogOnLoad 
setting info.header to 
There are 1 buttons
WEBSHELL- = 6
WEBSHELL- = 5
Move window by 118.5,115.6
screen x 17screen y 16
we have an event stuck -- removing it.
we have an event stuck -- removing it.
WEBSHELL- = 4
clearing PRIMARY clipboard
Document: Done (300.454 secs)
Error loading URL https://calendar-1.netscape.com/cgi-bin/nscal/login.cgi

Comment 1

[John Unruh]

Worksforme on two different Linux boxes. A network timing issue?

Comment 2

[sairuh (rarely reading bugmail)]

weird. i also tried 'http://calendar-1' as well as
'https://calendar-1.netscape.com.com/cgi-bin/nscal/login.cgi' --and i still get
the same, failing result.

Comment 3

[sairuh (rarely reading bugmail)]

arrgh, now using today's linux comm build, 2000.05.24.08, i'm unable to even run
the basic security smoketest --unable to access https://digitalid.verisign.com/.
it just spins and spins...and hangs if you try to interrupt it.

Comment 4

[Doug Turner (:dougt)]

I think that I see this also on windows.  can you confirm?

Comment 5

[John Unruh]

I am not seeing any SSL problems with today's WinNT or Linux builds, including 
https://digitalid.verisign.com/.

Comment 6

[sairuh (rarely reading bugmail)]

sigh. i'd try this out w/a new profile, but there have been nasty problems there
(bug 40294)... will doublecheck later.

Comment 7

[sairuh (rarely reading bugmail)]

another observation, fwiw: even if i click the unlocked icon (on an insecure
site), the browser will hang.

Comment 8

[Doug Turner (:dougt)]

This is probably a dup of http://bugzilla.mozilla.org/show_bug.cgi?id=40448

Comment 9

[sairuh (rarely reading bugmail)]

another data point: i trashed my .mozilla/, and with a semi-new profile (which
incurs bug 40211, not bug 40294 as i had thought), i still encounter this
problem.

Comment 10

[leger]

Putting on [dogfood+] radar.

Comment 11

[Simon Fraser [no longer active]]

dougt is away. this bug needs a new owner.

Comment 12

[Doug Turner (:dougt)]

well, I will be away from friday on.  I am still looking at this...

Comment 13

[Daniel (Leaf) Nunes]

this might go without saying, but in case it doesn't, a=leaf when you have a
fix.

Comment 14

[John Unruh]

I can't reach a secure site with Wed nights builds of Linux or WinNT.

Comment 15

[Doug Turner (:dougt)]

Try todays.  I checked in some changes last night that may fix this.  But, don't 
close this bug as we are still going to deadlock sometimes because of 40448.

Comment 16

[John Unruh]

The fix apparently didn't make it into this morning's LInux or WinNT builds.

Comment 17

[Doron Rosenberg (IBM)]

*** Bug 40436 has been marked as a duplicate of this bug. ***

Comment 18

[Doron Rosenberg (IBM)]

Still seen on mozilal nightly builds 200052508 win32 and linux.

Comment 19

[Daniel (Leaf) Nunes]

this is a blocker, but not holding changes for it. can we find an owner other
than doug, since he's going to be gone for the next month?

Comment 20

[Sitsofe Wheeler]

*** Bug 40653 has been marked as a duplicate of this bug. ***

Comment 21

[Terry Hayes]

Assigning to me.  I've determined that this is due to changes in the
PSMShimLayer code, which now leaves sockets non-blocking (causing errors in
PR_Connect to PSM) and uses timeouts for Recv (which causes operations like SDR
to timeout too early).  We need to review the requirements for these changes and
determine the correct action.

I'm working up changes to reestablish most SSL operations for Mozilla.

Comment 22

[Jim Roskind]

Terry: Since this is a dogfood bug, could you take a shot at an ETA, and put it 
in the status whiteboard?

Thanks,

Jim (dogfood slave driver in training) Roskind

Comment 23

[Terry Hayes]

I've checked in a temporary patch to back out two changes that were made in the
last couple weeks.  Need to verify that this fixes most problems, and determine
what the final solution should be.

Comment 24

[Doron Rosenberg (IBM)]

Accessing with 2000053108 Win32 Mozilla build without PSM isntalled will not
cause a crash anymore, but will give a "connection refused" error. However, I
can't check if it works with PSM, as the PSM at iPlanet just expired today.

Comment 25

[John Morrison]

I can load digitalid.verisign.com and the internal calendar-1 https servers
with 2000060108 linux. Will check win32 in a bit.

Comment 26

[stephen]

Still get ``connection refused'' going to secure sites with RH6.2 Linux and
Build ID 2000060308.

Comment 27

[John Unruh]

*** This bug has been marked as a duplicate of 41680 ***

Comment 28

[shrirang khanzode]

verif dup

Comment 29

[John Unruh]

Mass changing Security:Crypto to PSM

Comment 30

[John Unruh]

Mass changing Security:Crypto to PSM