Bug#881301: marked as pending

[Simon McVittie]

On Sat, 13 Jan 2018 at 20:38:39 +0900, Mike Hommey wrote:
> On Sat, Jan 13, 2018 at 11:35:08AM +0000, Simon McVittie wrote:
> > On Sat, 13 Jan 2018 at 08:50:14 +0900, Mike Hommey wrote:
> > > When can this reach unstable?

The source packages have now been accepted. The buildds haven't picked
them up yet, but you should see binaries later today.

> > Huh. Did you recently enable any new Shell extensions, or update
> > existing extensions, or anything like that?
> 
> I've had extensions enabled for a while, and reading this bug I disabled
> them all. That didn't stop.

As a general note, disabling an extension at runtime does not guarantee to
unload it fully (it's meant to, but bugs in the extension can mean that
its actions are not completely undone) so to be completely sure please
disable extensions, log out and log back in. (But if the Shell crashes,
that's also sufficient!)

> What /does/ seem to have stopped it is me killing my script that
> overwrites the background image every ~10 minutes. But I've been running
> that for well over a year.

Based on this, it's very likely that in the new version, you'll get
warnings logged (in the systemd journal if you're using systemd; I
don't know where they end up on non-systemd machines) with a JavaScript
stack trace. This will hopefully be more useful in tracking down the root
cause than the C backtraces we've already seen, which mostly just say
"I'm interpreting some JavaScript" without indicating what.

Please report each distinct stack trace as a separate bug, similar
to the one I just reported at
<https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=887082>.

Since you have a reproducer (your background-changing script) it would
be great if you could run without that script for a couple of hours,
report any stack traces you see with a note that they are *not* caused by
your background-changing script, then re-enable your script and report
any new stack traces that are well-correlated with background changes,
this time mentioning what your script does/how it works. There's probably
a bug in the code that sets the background.

Thanks,
    smcv