Re: [orca-list] Orca using too much memory on my machine?

[Mike Gorse <mgorse alum wpi edu>]

On Wed, 16 Jan 2013, Jason White wrote:

> If it were a C program I would suggest running it under Valgrind. However,
> it's a Python program, and I'm not sure what the equivalent tool is in this
> case. It could well be the AT-SPI 2 library that's leaking but you need to
> find out more precisely what the cause is.

If DEBUG_REF_COUNTS is defined in atspi-accessible.c, then it will count 
the number of accessibles that are cached and print out the count when an 
accessible is created or destroyed. I only use this to decide whether ref 
count leaks are present.

A while ago, I was poking around with gdb to figure out where the root 
accessible gets ref'd and figured out that pygobject leaks a ref when an 
object is accessed from a boxed structure (ie, when Orca asks for an 
event's source). I haven't taken the time to track down this leak so far, 
and it needs to be fixed in pygobject, so it is still there unless someone 
else fixed it recently.

That is just the leak that I know about. There may be others.

-Mike