:: Re: [DNG] couldn't register with ac…
Top Page
Delete this message
Reply to this message
Author: al3xu5 / dotcommon
To: dng
Subject: Re: [DNG] couldn't register with accessibility bus in Beowulf (workaround)
Wed, 25 Mar 2020 10:06:57 +0000 - Mark Hindley <mark@???>:

> I can conform that slim seems to do the right thing at the AT_SPI_BUS is set
> and available.
> On Wed, Mar 25, 2020 at 08:06:22AM +0000, Daniel Abrecht via Dng wrote:
> > Hi
> >
> > I've had that problem as well. [...]
> >
> > To me, it seams to be caused by a combination of something `lightdm` does,
> > and something `at-spi-bus-launcher` does not do.
> > After a login with `lightdm`, the X11 root window has the `AT_SPI_BUS`
> > property set (`xprop -root AT_SPI_BUS`), however, it seams the
> > `at-spi-bus-launcher`/`at-spi2-registryd` instance which created that
> > property does not exist anymore, causing applications to fail to connect to
> > it. In addition to this, the presence of that property prevents a new
> > instance of `at-spi-bus-launcher` from starting.

I had the same issue too after dist-upgrade from ASCII to Beowulf.

> > My current workaround is to remove the `AT_SPI_BUS` property, and then start
> > `at-spi-bus-launcher` again:
> > ```
> > xprop -root -remove AT_SPI_BUS
> > /usr/lib/at-spi2-core/at-spi-bus-launcher &

I have added:
export NO_AT_BRIDGE=1
in `/etc/environment` and also I have installed `at-spi2-core`.

This seems to have blocked the alerts, but I don't know if it's a real solution
or just a workaround that doesn't solve the underlying problem.

> Daniel, I get the same results as you with lightdm and your workaround is good
> too. Although I don't have to restart the at-spi-bus-launcher process
> manually, it is started automatically once a process tries to use it.

> The issue seems to be that lightdm starts its own AT_SPI_BUS which is then
> killed on login. But the AT_SPI_BUS property isn't removed so that it can be
> relaunched in the user's session. This is depsite Debian #760740 being marked
> as fixed.

Also I am using lightdm (+ MATE).

> That bug also suggests another workaround of setting 'xserver-share=false' in
> /etc/lightdm/lightdm.conf which also seems to work.

Maybe I'm wrong, but it seems to me that it is similar to the
`export NO_AT_BRIDGE=1` workaround. If so the 'xserver-share=false' setting in
the lightdm configuration seems to me to be more appropriate.



Say NO to copyright, patents, trademarks and any industrial design restrictions.
Public GPG/PGP key
ID:           4096 bit RSA key F94CFE23
Fingerprint:  59C6 9DC7 CD4B CF2F A190  E3DE 69C5 977B F94C FE23