366896 - Firefox crash after long runtime

Status: RESOLVED WORKSFORME

(Firefox :: General, defect)

Description

[w4zsvryct5gx7nf]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; hu; rv:1.8.1.1) Gecko/20061208 Firefox/2.0.0.1
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; hu; rv:1.8.1.1) Gecko/20061208 Firefox/2.0.0.1

After a week runtime with a lot of tabs/windows open (about 400), upon refreshing a page, it is opened in a new window, which has no controls (not even scrollbar). The original window hangs. Upon closing the new window, Firefox segfaults.


Reproducible: Always

Steps to Reproduce:
Well... it happens every time when I leave Firefox running for a longer time and have many open tabs/windows. But reproducing might require a week...

Comment 1

[Mats Palmgren (inactive)]

Sounds like a leak and that we eventually run out of memory...
How much memory does this PC have?
Please keep an eye on how much memory the 'firefox-bin' process consumes -
is it growing?  how much is it when it crashes?

Does it work better if you start Firefox in Safe Mode? (it temporarily disables
plugins and add-ons, this test will help us narrow down the cause of the
problem)  --  See http://kb.mozillazine.org/Safe_Mode

Comment 2

[w4zsvryct5gx7nf]

PC has 896MB memory. If a process is killed becuase the system runs out of memory there is an "OOMK" message in the system log, however when Firefox crashes there is no such log message.

firefox-bin consumes about 15MB shared, 45MB resident and 170MB mapped memory initially, before the crash it is usually around 40MB shared, 260MB resident and 340MB mapped.

I only use the Adblock Plus extension and no plugins; however I got the same crash once while not using any extensions as well.

Comment 3

[timeless]

please build firefox --enable-debug(ger-info-modules) --disable-optimize
and run w/ ulimit -c unlimited

Comment 4

[w4zsvryct5gx7nf]

OKay and I will post the coredump here when I get a crash next time.

Comment 5

[timeless]

not a core dump, just a stack trace from gdb from the crash

Comment 6

[w4zsvryct5gx7nf]

When I build Firefox from source, it's not reproducible (tried a lot). But with the official build it is. Are there any 'official' debug builds? I'm using gcc4.1.1 with Red Hat patches, maybe the stack guard pages prevent this from happening...

Comment 7

[Braden]

This sounds an awful lot like bug 263160; yet you claim you are getting a segfault. In my experience, closing the rogue window associated with bug 263160 causes the browser immediately to terminate abnormally--not crash.

Can you confirm that you are actually experiencing a segfault?

Comment 8

[Wayne Mery (:wsmwk)]

w4zsvryct5gx7nf ?

Comment 9

[Mats Palmgren (inactive)]

-> WORKSFORME