92348 - can't start mozilla with the same profile after crash

Status: RESOLVED FIXED

(Core :: Networking: Cache, defect)

Description

[Chun Che Li]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Windows; U; WinNT4.0; en-US; rv:0.9.1) Gecko/20010607
Netscape6/6.1b1
BuildID:    20010723

I can't start Mozilla 9.2 with the same profile after crash. I have to recreate
another profile through the profile manager to start mozilla.

Reproducible: Sometimes
Steps to Reproduce:
1. Start mozilla
2. Try to crash mozilla... I used the exit button(the cross) at the right upper
corner once. 
3. restart mozilla use the same user profile

Actual Results:  Mozilla broser doesn't show up. thought it appears in the task
manager. I have to recreate a new profile though the profile manager to start
Mozilla

Expected Results:  Well...  I want Mozilla to show up.

Comment 1

[Matthias Versen [:Matti]]

*** Bug 92349 has been marked as a duplicate of this bug. ***

Comment 2

[Matthias Versen [:Matti]]

1. is that only 0.9.2  or also with build 20010723 ?
2. Are you using the -turbo option ?
3. You wrote:"I used the exit button(the cross) at the right upper corner once."
   Dou you mean you crash if you use the "x" ?
4. Have you installed Mozilla in a fresh directory ? (not over an older build ?)
5. Have you the writer permisssion for your profile directory ?
6. After a crash have you tried to logoff/logon and try it again ? (file locking 
   problem)
 

Comment 3

[basic]

sounds like something in the profile is messed up. What did you do to cause the
crash? Did you make copies of bookmarks many times? (weird question I know, but
I've seen this after doing that before)?

Comment 4

[Chun Che Li]

Here is one way to reproduce the buy on  Mozilla 20010726
1) Create a new profile by profile manager
2) change to Memory and disk cache to 0
(I did this by Edit-> perfencece->advance -> cache and set the value = 0)
3) exit the mozilla (if this doesn't  reproduce the bug, try use task manager
and kill mozilla)
4) restart mozilla with the same profile. nothing shows up.

However, the same profile works for Netscape 6.1  

Comment 5

[Doron Rosenberg (IBM)]

you should never use the same profile with ns6.x and mozilla by the way...

Comment 6

[Matthias Versen [:Matti]]

This sounds like : Networking:Cache

gordon: Is this a known cache problem ?

Comment 7

[Peter Trudelle]

Does anyone know if QuickLaunch (indicated by pref, icon in systray) is
involved?  This can also be triggered by the -turbo command line switch.

Comment 8

[gordon]

If both the disk and memory caches are disabled we still return a cache entry
descriptor when a client calls OpenCacheEntry(), because we lazily bind an entry
to a specific device.  If neither device is enabled, we fail when operations are
performed on the descriptor which would cause it to be bound to a device.  The
imgLib code is ignoring some of these errors.

I think it would make more sense to fail earlier, and not return a cache entry
descriptor if we have no cache devices enabled.  Adding this one line
accomplishes that and enables the browser to come up:

	if (!mEnableDiskDevice && !mEnableMemoryDevice)  return NS_ERROR_FAILURE;

Comment 9

[gordon]

Attachment: Patch causing OpenCacheEntry() to fail if no cache devices are enabled.

Comment 10

[gordon]

Pav, Gagan, I'm looking for an r=.  Darin, I need an sr=.

I also think we should fix this on the branch.  Nominating for nsbranch.  Gagan,
how about a +?

Comment 11

[gordon]

Attachment: Added check for stream based cache entries.

Comment 12

[Gagan]

Comment on attachment 49957 [details] [diff] [review]
Added check for stream based cache entries.

r=gagan

Comment 13

[Gagan]

this allows us to restart the browser after setting disk/mem size to zero,
without problems! so approving for branch

Comment 14

[gordon]

Attachment: Corrected logic for non-stream-based check.

Comment 15

[Doug Turner (:dougt)]

good catch gordon.  sr=me.

Comment 16

[Darin Fisher]

Comment on attachment 49963 [details] [diff] [review]
Corrected logic for non-stream-based check.

sr=darin (are there other entry points that need to be similarly protected?)

Comment 17

[gordon]

No, the creation of all descriptors goes through
nsCacheService::ActivateEntry(), unless the cache entry is already active.

Comment 18

[gordon]

The patch has been checked into the trunk.  Leaving bug open in hopes of getting
PDT approval tomorrow to land on the branch.

Comment 19

[Jaime Rodriguez, Jr.]

Pls come to PDT @ 3:30 today to talk about this one.

Comment 20

[Jaime Rodriguez, Jr.]

check it in tomorrow afternoon, after we get some testing on the trunk - PDT+.
Pls spam people to make sure people test it out.

Comment 21

[gordon]

Fix checked into branch.  Marking FIXED.

Comment 22

[Tom Everingham]

set mem and disk cache to 0, killed moz, restarted fine.

verified on branch only:

WinNT 2001-09-24-05-0.9.4

Comment 23

[benc]

.capacity=0 should shut the specific cache service off (making it not show up in
about:config).

VERIFIED: Mach-O, mozilla 1.4b