Memory leak in the server

[corrados]

I just observed the following on the Central Server: Right now we have heavy traffic on that server. The server run for about a month and the Jamulus server instance used about 500 MB. When I reboot the server, Jamulus just consumes about 4 MB.

I think the following code might be the problem:

jamulus/src/serverlogging.cpp

Line 65 in ab785cb

QTextStream& tsConsoleStream = *( ( new ConsoleWriterFactory() )->get() );

We do a "new" all the time but do not delete that instance. Maybe it is better to completely remove the ConsoleWriterFactory() class and use QInfo instead.

[pljones]

Yes, I'd back that, as we've moved to a Qt version dependency where qInfo() will always exist. It'll need a little refactoring (as the messages format differently).

[corrados]

Would you like to do it or shall I do this change?

[pljones]

👍 I'll do it now - shouldn't take long.

[corrados]

Thanks.

[pljones]

Pull request: #926

[hoffie]

@pljones Now that #926 is merged, this could be closed?

[softins]

@pljones Now that #926 is merged, this could be closed?

If indeed that was the source of the leak. Time will tell!

[pljones]

Yes, it's probably wise to monitor a server - preferably one that's long running and heavily used - to see if this the problem has gone away.

[pljones]

Can we tag this as needing some kind of post-3.7.0 review (and hopefully then closing)? I'd expect checking the process size on the Any Genre 1 directory server immediately after installing 3.7.0, then at intervals of a week for three weeks should provide enough information. If the size is stable, it's fixed.

[genesisproject2020]

I compiled the code yesterday (3.6.2dev-6d79bf1b) and it has now been running on "Swedish Open Jam 1" for 25h. During that time the amount of allocated memory went from 40Mb to, at the moment, 1Gb. I've also changed it on a private instance where a group had a rehearsal this evening with 16 members and after the 2h rehearsal the service have allocated 4,3 Gb of memory (see picture) and it is stuck there now when the instance is idle again. I do believe something is still eating the memory. Can I assist in someway to help finding the leak?

My environment: Ubuntu 20.04 LTS (5.4.0-65-lowlatency)