[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: gEDA-user: PCB GL Memory leak

On Sat, 2011-05-14 at 23:53 +0100, Thomas Oldbury wrote:
> The "pcb" process under System Monitor shows usage. It seems to
>    allocate it in large blocks of ~100MB at a time and never lets go of
>    these.

Are you working on a particularly complex board?

I've tried git HEAD PCB and nothing jumps out at me in terms of memory
leaks. I note your 2D xorg driver is a little old at 2.9.1, current is
now 2.15.0

I recall there were some problems in the past with graphics resources
not being freed correctly (might have been kernel related too), so if
you are consistently seeing a leak with simple boards, upgrading your
kernel and driver stack might well fix it.

FWIW, when I last had this problem last, the system monitor applet (the
one which draws little graphs in your gnome-panel), the memory usage was
coloured as cache, rather than resident memory. The bug caused the
system to start swapping when it ran out of memory, despite it thinking
that the memory was cached data, and as such could have been dropped.

If you have the performance monitoring applet (I forget the exact name),
it would be worth tracking which colour the memory climb in PCB's usage
is - either the colour for active memory usage, or cached pages.

(I'm happy to try and reproduce leaks if you can give me good
instructions to do so, ideally with an example board we can both use).

HOWEVER: PCB keeps a lot of undo data. On a highly complex 8 layer board
test-case I have, I can easily get to 300M resident just loading the
board, with bumps of say 20Meg for every power plane deleted and

Obviously this isn't ideal, but it could possibly explain where your
memory usage is going.

Peter Clifton

Electrical Engineering Division,
Engineering Department,
University of Cambridge,
9, JJ Thomson Avenue,

Tel: +44 (0)7729 980173 - (No signal in the lab!)
Tel: +44 (0)1223 748328 - (Shared lab phone, ask for me)

Attachment: signature.asc
Description: This is a digitally signed message part

geda-user mailing list