Re: Bug#953971: Please add at-spi2-core to the build-dependencies so the tests on hurd-i386 can succeed

[Guido Günther <agx@sigxcpu.org>]

Hi Samuel,
On Wed, Mar 25, 2020 at 01:59:50AM +0100, Samuel Thibault wrote:
> Hello,
> 
> I got further down with this (I happen to be maintainer of the
> at-spi2-core package :) ), the difference here is that the dbus-x11
> package is not installed in the Linux case. If I install the dbus-x11
> package, Linux starts getting the same issue: /usr/bin/dbus-launch gets
> launched to autostart a session dbus bus and libatspi then tries to find
> its at-spi2-core service, to no avail. Normally this only produces a
> glib warning, but apparently the tests here enable making glib warnings
> fatal, thus the trap.

Yes, one of the purposes is to test for runtime errors so we trap on
warnings and criticals.

> So I was wondering whether it'd be a bug of the hurd port or of
> libatspi, but apparently no, it's just an unfortunate combinaison of
> behavior.
> 
> Installing at-spi2-core or setting NO_AT_BRIDGE=1 should be fine then.
> I would say it would be safer to do so on non-hurd ports too, in case
> at some point the dbus packages end up making dbus-launch available by
> default on non-hurd too.

Thanks a lot for investigating. I'll add NO_AT_BRIDGE=1 to the next
upload. Cheers,
 -- Guido

> 
> Samuel
>