Re: [LAU] qtractor crashes (was: the defacto host for lv2)

From: Rui Nuno Capela <rncbc@email-addr-hidden>
Date: Sat Jun 05 2010 - 23:46:39 EEST

On 06/05/2010 07:19 PM, Niels Mayer wrote:
> Potentially, multithread access to shared memory data-structures seems
> to be the cause of hard-to-reproduce error messages coming out of
> Qt-based apps (see also https://bugs.kde.org/show_bug.cgi?id=240391 ):
>
> gnulem-109-~/Documents> qtractor
> *** glibc detected *** qtractor: corrupted double-linked list:
> 0x00000000012ab6a0 ***
> ^C
>

this is most likely one hideous memory corruption bug, runaway memwrite,
double or even triple memleak. whatever.

do not (ever) rule out my sloppy coding habits.

> You may never see these errors until you run these apps on a fast
> quad-core machine. For Qtractor -- my hunch of the day -- check for
> simultaneous access to the "inputs" and "outputs" configurations;
> after loading several projects or activating "track->inputs"
> "track->outputs" or the "outputs" and "inputs" buttons in the mixer
> (all of which pop up the "Connections" window)... do this a few times
> and eventually your midi devices disappear in the gui. You have to go
> to the "Connections" window's option-menu and select "(All)" or
> reselect the current option in order to refresh the display.
> Something's getting "clobbered" in the midi or audio connections
> listings each time you load a new .qtr or .qtt file... and this is an
> error that recurs more quickly on a fast quad-core than a slower
> dual-core.
>

i do have a quad, where is my primary devel host, intel q8200 though,
maybe not the fastest, but still find some of those reported crashes
very hard to reproduce.

> ... makes me think i was right on track with all my comments about
> clojure, garbage-collection, and software transactional management :-)
> ... i think i'll need a supercomputer to run "valgrind --tool=helgrind
> qtractor" ...
>

indeed. valgrind may help but it may get you also bored to death even on
the fastest of the machines, unless you can reproduce the crashes in a
finite, consistent and determined number of steps. i've been there i've
done that and still chose to step over and run forward :)

cheers

-- 
rncbc aka Rui Nuno Capela
rncbc@email-addr-hidden
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user@email-addr-hidden
http://lists.linuxaudio.org/listinfo/linux-audio-user
Received on Sun Jun 6 00:15:04 2010

This archive was generated by hypermail 2.1.8 : Sun Jun 06 2010 - 00:15:04 EEST