[Tizen General] Does GBS leak memory?
Zhang, Qiang Z
qiang.z.zhang at intel.com
Fri Sep 6 16:28:21 GMT 2013
It's not memory leak, gbs local full build is really memory consume task, especially for some large project, such as kernel/webkit.
There're some discussion before, for webkit build at least 8G (4G memory+ 4G swap) needed, so if you have memory less than 8G, you may got "OOM" error even you just build one webkit package.
Actually, you will get the same error by using rpmbuild instead of gbs. I remember some engineers suggest to split webkit into several small packages, not sure the result now.
The recommended memory is 8G physcal memory and 8G swap.
From: general-bounces at lists.tizen.org [mailto:general-bounces at lists.tizen.org] On Behalf Of Hanchett, Paul
Sent: Friday, September 06, 2013 11:44 PM
To: Ding, Jian-feng
Cc: Kenneth Smith; general at lists.tizen.org; Steven Maurer
Subject: Re: [Tizen General] Does GBS leak memory?
Well, I definitely don't have 6GB to give on my current dev machine-- limited to 2GB for my VM before the (Windows) host becomes starved for memory. I'm told my real dev machine is "in the mail". ;-)
This list won't let me post an image, but the gbs job has exited and top shows a lot of memory still consumed (1.9GB real memory and 1GB swap on a machine with 2GB main and 1GB swap disk). The particular image I have shows cc1 still running, but I've seen where everything has exited and it still looks like large amounts of memory are consumed (~1.7GB!). I can only account for maybe half that as Ubuntu...
The other thing that may be significant is that it only seems to happen when it fails because of out of memory (I have seen a message about failed to fork...). Doesn't happen on a "normal" error.
I seem to have gotten past this for now by removing the "--threads=4" parameter from my build line.
MSX on behalf of Jaguar Land Rover
One World Trade Center, 121 Southwest Salmon Street, 11th Floor, Portland, Oregon, 97204
Email: phanchet at jaguarlandrover.com<mailto:phanchet at jaguarlandrover.com>
Jaguar Land Rover Limited
Registered Office: Abbey Road, Whitley, Coventry CV3 4LF
Registered in England No: 1672070
On Fri, Sep 6, 2013 at 4:51 AM, Jian-feng Ding <jian-feng.ding at intel.com<mailto:jian-feng.ding at intel.com>> wrote:
Actually, the gbs bulid task is quite a memory consuming process, as we all
know. So in our testing machines, for the local full build, we usually need
more than 6GB physical memory, especially for several huge projects, like wrt.
But the mem leaking, I don't think it will exist. gbs will call some perl
scripts inside, as separated child-processes, which will not leave memory
unreleased as I understand.
Our developers will check it more, and will provide more information if found
On Wed, Sep 04, 2013 at 11:45:16AM -0400, Hanchett, Paul wrote:
> I just attempted a gbs build which ran for hours, fairly successfully,
> until it terminated with an "out of memory error".
> Top shows 2GB of memory available and around 1.7GB in use (and some amount
> of swap that I didn't note) even with all windows and processes closed
> except for the shell running top.
> Looking at memory usage doesn't show all that much allocated to any single
> Rebooting the VM and running top shows ~783MB in use and no swap. (I find
> even that much sort of surprising; ubuntu unity must not be all that
> Am I just misreading the output of top, or is there really a memory leak?
> TIA for any help!
> Paul Hanchett
> Infotainment Engineer
> MSX on behalf of Jaguar Land Rover
> One World Trade Center, 121 Southwest Salmon Street, 11th Floor, Portland,
> Oregon, 97204
> Email: phanchet at jaguarlandrover.com<mailto:phanchet at jaguarlandrover.com>
> Business Details:
> Jaguar Land Rover Limited
> Registered Office: Abbey Road, Whitley, Coventry CV3 4LF
> Registered in England No: 1672070
> General mailing list
> General at lists.tizen.org<mailto:General at lists.tizen.org>
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the General