lubuntu-qa team mailing list archive

[Lars Noodén <lars.nooden@xxxxxxxxx>]

On 01/04/2013 09:54 PM, Chad MILLER wrote:
> On Thu 03 Jan 2013 06:49:47 PM EST, Phill Whiteside wrote:
>> We are seeing an issue with Chromium on low RAM systems where we get "It's dead Jim" issues. At first I believed this to be a Chromium memory leak, but it does not affect all web pages [1]. I have tried to report this via [2] which errorred out with 'mal formed request', trying to use the 'learn more' and trying to report an error in the failed window also resulted in the same. Even more odd (to me) was that leaving the tabs open ended up in them giving the "It's dead Jim" error on the tab. Just as a check in my own logic, I opened up a launchpad bug tab for a bug. This tab, over several days & resets of other tabs that had to be reset never once failed.
>>
>> Apologies for the long introduction. Is there anything I can add to the system to try and get some details for you guys to work off? gdb I have been intimated at, is not really suited for a memory leak; but as I don't think it is a memory leak and Chromium still runs with the tabs still 'alive' would it be of use to use gdb to trace back an PiD?
>>
>> Assuming you don't have a machine that you can pull memory chips out of, until you have 512Mb of RAM (The guys with these machines are the ones who 1st reported it)..
>>
>> 1. Create a VM with 512 Mb RAM
>> 2. Install Lubuntu (I suggest using alternate at such low RAM) [3] - I've also tried this with Raring
>> 3. Open a couple of tabs, e.g. BBC News [4] and a bug report [5] - Well, it was a chromium bug :)
>> 4. Open a couple of other tabs for sites you know to be stable - Remember, you are on a low-RAM system, nothing too exotic!
>> 5. Wait.
>> 6. Tabs will report "It's dead Jim".. - This may take several hours - Tab opened with [5] will stay working.
>>
>> Us testers are stuck to try to progress and any help you can give to help log the bug correctly in order that it can be progressed is needed. I've tried installing the dev chromium instead of the 'new' chromium in the repos, the effect is the same.
> 
> 
> Hi Phill, all.
> 
> I assume it's the out-of-memory process reaper in the kernel, at work 
> here killing process that backs the tab you see.  Can you confirm 
> something interesting in "dmesg" output?
> 
> I can imagine that a lazily-written web site can have Javascript code 
> that ever grows its memory usage.  I'm keen to know whether the same 
> machine can reproduce this crash on a mundane web site that doesn't 
> have JS events firing off RPC calls and/or updating the DOM.  If it 
> does crash anyway, this gets interesting.
> 
> Answering those two questions would help me categorize this bug-report 
> pretty easily.  Let's open a normal bug on Launchpad to track this for 
> now.
> 
> - chad
> 
I see the out-of-memory "it's dead" as possibly something different.
But it does happen on pages without javascript.  Usually it takes a few
hours of running to start to get the "it's dead" errors.  It's pretty
easy to reproduce, just use chromium for 5 or 6 hours running.

Regards
/Lars