243540 - Testcase (exploit ?) leaks 2MB/sec and hang Mozilla

Status: RESOLVED WORKSFORME

(SeaMonkey :: General, defect)

Description

[Olivier Cahagne]

Build ID: 20040513 on Win2k & 20040512 on Linux.

Steps to reproduce:
1. Load testcase,
2. Mozilla memory usage grows 2MB/sec and CPU is at 99%

Comment 1

[Olivier Cahagne]

Attachment: Testcase (warning: will maxout CPU and RAM quickly)

Comment 2

[tor]

while(true) JavaScript running unbounded.

Comment 3

[Boris Zbarsky [:bzbarsky]]

I wonder why the branch callback is not catching this...

Comment 4

[Nicholas Allen]

The callback will actually kill it if you wait long enough.  It's spending an 
enormous amount of time in editor+txmgr+layout each time it goes out from js.  
There's less than 100 callbacks/second in an optimized build on a fast 
machine.  We maybegc every 4k callbacks and then cut it off at 64k callbacks 
which is the first time the callback timer is checked.

Comment 5

[tenthumbs]

It's the embedded Ctrl-U that's making mozilla eat memory. Remove it and mozilla
just spins for a while before it asks it it should kill the script.

Comment 6

[Brendan Eich [:brendan]]

So back to browser-general?  This was never JS engine.

/be

Comment 7

[Olivier Cahagne]

Script is caught by the warning 'Unresponsive script' on both Camino 0.9a2 and
Firefox Deep Park alpha 2.
Marking WFM.