Bug#940018: at-spi2-atk autopkgtest fail most of the time on s390x

[Samuel Thibault]

Sebastien Bacher, le jeu. 12 sept. 2019 14:17:40 +0200, a ecrit:
> Le 12/09/2019 à 13:23, Samuel Thibault a écrit :
> > But then the problem is that testbed-packages does not contain
> > at-spi2-core. The tests can't work without it.
> 
> Samuel, I think you are chassing the wrong thing there. The tests are
> working on other archs and hit a SIGTRAP on s390x,

They are not working on arm64 either.

I have tried to run on the arm64 and s390x Debian porter boxes in a
dchroot and it succeeded there. I wanted to make sure that I have the
same set of packages.

Now that the list of installed packages is clear and does include all
that is is needed, I'd like to know the environment variables. Normally
the porterbox dchroots are quite raw, but who knows.

Also, do autopkgtest run with a controlling terminal or not? That might
make a difference, and I have always run it in a terminal only.

> also the environment is what is installed on the system by default

Which system?

> (which is what the testbed artifacts lists) + the 'installed packages
> list' I copied earlier,

That "+" was not obvious, and probably needs to be made more obvious in
the artifact names.

> and that includes at-spi. I guess it would be useful to get a
> backtrace from the sigtrap, 

The problem is that from what I see, it's not the test which crashes,
but one of dbus-daemon/at-spi-registryd/at-spi-bus-launcher.

> I will see if I can get access to an similar env to reproduce/get one

That's what I was trying to do.

Samuel