Assistive Technologies is spamming .cache with many folders like at-spi2-AYBRGW for example

Bug #1020512 reported by philinux
36
This bug affects 8 people
Affects Status Importance Assigned to Milestone
at-spi2-atk (Ubuntu)
Fix Released
Undecided
Unassigned

Bug Description

Two processes are running, at-spi2-registryd and at-spi-bus-launcher, even though I've unticked AT-SPI D-Bus instartup apps.

I've uninstalled all but at-spi2-core and the spamming continues. Even after killing the above to processes. So I'm guessing dbus must be involved. Please feel free to change the package for the bug.

The contents of all the folders contain 1 file named socket of 0 bytes. This file cant be opened.

ProblemType: Bug
DistroRelease: Ubuntu 12.10
Package: at-spi2-core 2.5.3-1
ProcVersionSignature: Ubuntu 3.5.0-3.3-generic 3.5.0-rc5
Uname: Linux 3.5.0-3-generic x86_64
ApportVersion: 2.2.5-0ubuntu2
Architecture: amd64
Date: Tue Jul 3 12:14:29 2012
InstallationMedia: Ubuntu 12.10 "Quantal Quetzal" - Alpha amd64 (20120609.1)
ProcEnviron:
 LANGUAGE=en_GB:en
 TERM=xterm
 PATH=(custom, no user)
 LANG=en_GB.UTF-8
 SHELL=/bin/bash
SourcePackage: at-spi2-core
UpgradeStatus: No upgrade log present (probably fresh install)
mtime.conffile..etc.xdg.autostart.at.spi.dbus.bus.desktop: 2012-06-25T13:25:47.813621

Revision history for this message
philinux (philcb) wrote :
dino99 (9d9)
tags: added: i386
Revision history for this message
dino99 (9d9) wrote :

Here on Quantal i386, i get the same issue: lot of .cache/at-spi2xxxx/ subfolders with a single file -> socket=0 byte about network.
Look like they are created while opening a session (gnome-classic here). Wonder if its related to the fact that i need to restart the networking service, after adding a nameserver into resolv.conf, or not. But i beleive this flood have started after the libatk-bridge2.0-0 2.5.6-0ubuntu1

Revision history for this message
Launchpad Janitor (janitor) wrote :

Status changed to 'Confirmed' because the bug affects multiple users.

Changed in at-spi2-core (Ubuntu):
status: New → Confirmed
Revision history for this message
Elfy (elfy) wrote :

Same thing appears to be here as well with Xubuntu Quantal

Revision history for this message
howefield (howefield) wrote :

Doesn't seem to affect Unity 2D.

Revision history for this message
Nazar Mokrynskyi (nazar-pc) wrote :

The same on Ubuntu Quantal x64

Revision history for this message
Luke Yelavich (themuso) wrote : Re: [Bug 1020512] Re: Assistive Technologies is spamming .cache with many folders like at-spi2-AYBRGW for example

GNOME upstream for 3.5/3.6 have decided to enable accessibility by default, which means at-spi2 is always going to be running, hense why trying to disable it from startup applications does nothing. It needs to be hidden from startup applications, but the startup file is still needed for other desktop environments which still enable accessibility selectively.

Is this adversely affecting the performance of your system? I get the same thing here on Quantal, and all those at-spi directories only take up 1.7MB or so. If you are not using orca or other assistive technology tools, then you shouldn't notice any difference.

Revision history for this message
Doug McMahon (mc3man) wrote :

Luke -
I'm cdrtainly not yet in a position to say whether there are any perf issues, like many others there is enough else gone south to preclude judging at all
(here nautilus has a half-life of 15 sec's or so

One 'issue' may be is at what point, if any, do these unique dir.'s stop being created?
After 1 hr. with one restart & a couple of log out/in's I've got 120 dir.

Will I have 500 or more next week, 3000 next month, ect., ect.?.

Atm users can stop this by removing at-spi2-core, whether advisable don't know, whether there are plans to dep at-spi2-core other than to ubuntu-desktop also don't know

Revision history for this message
Mgorse (mgorse) wrote :

These are directories containing sockets that at-spi2-atk and libatspi use for direct D-Bus connections. The sockets used to be placed in /tmp/at-spi2. This was changed in 2.5.3 so that they now live in XDG_RUNTIME_DIR (see https://bugzilla.gnome.org/show_bug.cgi?id=678348). I've randomized the directory name at the request of a couple people who reviewed my original patch, and now a separate directory is created for each application, and part of the issue is that gtk+ does not currently call the clean-up code in atk-bridge, so nothing removes the sockets and directories.

I've just committed a change to at-spi2-atk (6fbb1b) to make it lazily create the directory and socket (ie, only when an AT actually requests the socket's address), so it should no longer create directories if no AT is listening. The change will be included in 2.5.4 (I'll release on the 16th). Note that this does not actually fix the problem--it only hides it from users who do not make use of AT-SPI.

Revision history for this message
philinux (philcb) wrote :

I've gone into /usr/lib/at-spi2-core and rename the two files at-spi2-registryd and at-spi-bus-launcher as .backup.

Temporary work around.

Revision history for this message
dino99 (9d9) wrote :

This update seems to stop the creation of these sockets folders:

at-spi2-atk (2.5.4-0ubuntu1) quantal; urgency=low

  * New upstream release
    - Fix for bug #679013 - AtspiCollection should be implemented for
      all containers
    - Don't create a (non-readable) subdirectory for the socket when root.
    - Plug ref count leaks (BGO#679285).
    - Only create a directory and a socket when requested. (LP: #1020512)

Waiting some more time to know how much its fixed.

Revision history for this message
dino99 (9d9) wrote :

Seems to be fixed now here on Quantal i386 & logged as gnome-classic.

Luke Yelavich (themuso)
affects: at-spi2-core (Ubuntu) → at-spi2-atk (Ubuntu)
Changed in at-spi2-atk (Ubuntu):
status: Confirmed → Fix Released
Revision history for this message
kenn (whatnext) wrote :

I have the same problem on Ubuntu 14.04.3

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.