Redshift does not quit upon logout from session

[mtvoid]

Redshift does not seem to be managed by the desktop session. So, when gtk-redshift is launched upon login to the desktop, it spawns a redshfit process, but it is not terminated when logging out. Logging in again launches a new redshift process; doing this a few times will lead to multiple stray redshift processes hanging around (you can see it by 'pidof redshift'). This is clearly not the correct behaviour, there should only be one instance of redshift running, and it should die when the session exits.

[Paul Natsuo Kishimoto (khaeru)]

I can confirm this.

[Launchpad Janitor (janitor)]

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

[David F. (malteworld)]

The problem arises because upon session end gtk-redshift receives SIGTERM, which prevents python from running the "finally" section. The included patch registers a signal handler, which calls sys.exit(), which in turn leads to python executing the "finally" section, where the child process is terminated.

[David F. (malteworld)]

Unfortunately, the previous method is not optimal because SIGKILL cannot be intercepted. One could however call prctl.set_pdeathsig() (analogous to PR_SET_PDEATHSIG of prctl(2)) after forking in the child process.

[Ubuntu Foundations Team Bug Bot (crichton)]

The attachment "Intercept and forward SIGTERM" of this bug report has been identified as being a patch. The ubuntu-reviewers team has been subscribed to the bug report so that they can review the patch. In the event that this is in fact not a patch you can resolve this situation by removing the tag 'patch' from the bug report and editing the attachment so that it is not flagged as a patch. Additionally, if you are member of the ubuntu-reviewers team please also unsubscribe the team from this bug report.

[This is an automated message performed by a Launchpad user owned by Brian Murray. Please contact him regarding any issues with the action taken in this bug report.]

[Ryan S Kingsbury (ryanskingsbury)]

Confirmed on Fedora x64 with Gnome 3.4.

[ricardofunke (ricardo-funke)]

Hi,

I'm affected by this bug and I see that it's kinda old. So does redshift is discontinued? Any chance of solving it?

I'm using Gnome 3 and I tryed some workarounds wighout success like .bash_logout, or /etc/gdm/PostSession/Default.

Does you have some other tips for me?

Thanks in advance and best regards.

[David F. (malteworld)]

I wrote the author/maintainer a couple of months ago to see, if he's still active. No response so far.
That leaves us (and me) the possibility to fork redshift into a PPA.

For the time being you can apply either of my patches (preferably the latter) to /usr/share/pyshared/gtk_redshift/statusicon.py, if you don't want to rebuild from source.

[ricardofunke (ricardo-funke)]

Hi David,

I applied the patch "Call prctl.set_pdeathsig() after forking" but it still doesn't kill the process at logout.

After applied, I removed statusicon.pyc and statusicon.pyo and then ran "python2 -m compileall statusicon.py" to regenerate.

But the problem persists, I log out from my account and log in into my wive's account, there I still can see my redshift process running with ps command, the only process remaining from my user.

[David F. (malteworld)]

Hm, the patch was definitely applied correctly and python would usually recompile outdated object files.
You probably did, but I'm asking just to make sure: did you terminate the old remaining processes "redshift" and "gtk-redshift", and install "python-prctl", the dependency introduced by the patch? Does gtk-redshift spill any interesting messages when run from command line?

[Jon Lund Steffensen (jonls)]

I've been working on an update for gtk-redshift to be able to provide more information to the user. That has meant that I've had to rewrite most of the process management code and for that reason I'm not sure how this discussion applies any more. If you want to take a look the code is here: https://github.com/jonls/redshift/tree/interactive-gui . A sigterm handler is installed and it seems to take care of shutting down the child process properly. However, even with a SIGKILL the child process seems to be shut down seemingly contrary to what David said in #4.

[deutrino (deutrino)]

This is still a problem. Who is the maintainer that can review the patch(es)? Downgrading to a PPA isn't great.

[deutrino (deutrino)]

Per http://jonls.dk/redshift/

"The project is managed at GitHub where issues and pull requests can also be posted. Please don’t use the old bug tracker at the Launchpad project page anymore."