Bug#924496: 'realloc(): invalid next size: 0x000055a779ef2170' crash when opening iPod w/ ~12000 tracks

[Fred Korz]

Hello Bernhard,

Now it (a) loads completely without crash from the same iPod (and no
changes there), and (b) does so in <50% as long.

Arrgh! I hate Heisenbugs!!!!  It was entirely repeatable last week, 3 for 3.

I've not rebooted since before my report, nor has the rhythmbox package
changed version (3.4.3-2) ,
but any of the dependencies may have been updated by automation.

I installed the debugging symbol packages, then started under gdb, plugged
in the iPod and no load-up crash, plays fine.
I then ejected and ran rhythmbox without gdb, plugged in the iPod, and
again no load-up crash, plays fine.

Some more answers embedded below.

On Tue, Mar 19, 2019 at 10:27 AM Bernhard Übelacker <bernhardu at mailbox.org>
wrote:

> Hello Fred Korz,
> I just tried to get some more information out of backtrace,
> without having an iPod or being involved on packaging rhythmbox...
>
> But am I right this "Debian Release: rodete" is a version
> of gLinux - Googles internal rebuild of Debian testing?
> Can this be downloaded somewhere?
>

The name, "rodete", is "ROlling DEbian TEsting" and apparently a pun in
spanish as well.
It is Debian testing but, as I understand it, run through an internal
"sieve" of tests before rolling out a consistent snapshot to users.
It's sort of what would happen if one lagged testing by about 1-2 weeks,
though some packages can be closer to testing's head if urgent.


> And are there debug symbols available for installation?
>

Yes they are.  I've installed these debug symbol packages at work and will
install at home tonight.


> In Debian these packages are available in a separate
> repository [1] and are named like this:
>
>     rhythmbox-dbgsym librhythmbox-core10-dbgsym libglib2.0-0-dbgsym
> libtdb1-dbgsym
>
> If yes, you could try to install them and run rhythmbox
> like this and provide the output:
>
>     gdb -q -ex 'set width 0' -ex 'set pagination off' -ex 'run' -ex 'bt'
> -ex 'detach' -ex 'quit' --args /usr/bin/rhythmbox
>

Damn Heisenbug.  Blew out 3 times last week, once with a coworker there to
see it.  None this time, either with gdb or without.


>
> As this fault seems to be inside the memory allocator, maybe
> setting "export MALLOC_CHECK_=2" might reveal some more details?
>
> Can this fault be reproduced on a plain Debian testing, too?
>

I'll try tonight/tomorrow (20190319/20190320) on a system at home where
I've been running Debian testing for 14+ years now,
usually update nightly, and rarely get burned by something slipping through
from experimental into testing that wasn't quite ready.

It's likely that I'll have to install rhythmbox + symbol packages. I've had
no need for rhythmbox there.  That system is where the backup
copy of my music library lives and I use vlc directly from the files, or
serve my library via forked-daapd (successor to firefly / mt-daapd).


> Kind regards,
> Bernhard
>

Thanks for the guidance.  I wIll both (a) get back to you with results -
reproduction or Heisenbug - and (b) keep the instructions in case
of some future return of the Heisenbug, hoping to get a better capture.


> [1]
> https://wiki.debian.org/HowToGetABacktrace#Installing_the_debugging_symbols
> [1]
> https://stackoverflow.com/questions/6750815/how-to-turn-off-glibc-run-time-protections
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://alioth-lists.debian.net/pipermail/pkg-gnome-maintainers/attachments/20190319/144843dd/attachment.html>