[mythtv-users] mythbackend still eats memory: the current status

[Udo van den Heuvel]

Alex Butcher wrote:
> It could well be that your limited hardware resources and/or usage pattern
> accentuate such leaks to the point where they are more easily observable
> over shorter time periods. 

What do you mean with this?
Normally the backend runs OK until I restart it.

> Certainly one of the difficulties I've had in
> analysing the data I was collecting was having the mythbackend process run
> for long enough to be meaningful - what with reboots, power outages,
> periodic rescanning of multiplexes with mythtv-setup, the two months
> continuous uptime is somewhat unsual, and the 75MB growth in VSZ over that
> period didn't present a practical usability problem for me.

It is not so much a usability issue for me, others had more bad luck 
(see the mailinglist, they had to reboot sooner), but rather and at 
least a quality issue. Of course I wonder at the numbers and decide 
after a few weeks to restart the backend (if not upgrade to a new svn in 
the same go).
And of course if I wait much longer I will get usability issues (swap is 
limited to 1G or so).
The key is that if memory leaks are still happening while doing the most 
basic task that the software was written for, what other issues might be 
lurking around the corner? I want to know and if possible pinpoint them.
In this case the hardware limits me a bit, or perhaps valgrind needs to 
much hardware.

So the consensus is to use valgrind where possible. When a leak is 
found, post the logs here...

Kind regards,
Udo