dunst - a dmenu-ish notification system

arclance · 2012-07-16 03:00:09

^ It works fine after I made my own package that allowed me to remove notify-osd.

I did see a huge (~200 MiB/min) memory leak in dbus-daemon during the few seconds they were both installed.

My problem was that "mail-notification-evolution" depended on "notification-daemon".

Since this is not even a runtime dependency but a functionality dependency you can replace the notification program with whatever you want.

As long as its package has "Provides: notification-daemon" the "mail-notification-evolution" package will stay installed.

The only reason notify-osd was installed on that computer was because "mail-notification-evolution" needed something installed to satisfy its "notification-daemon" dependency.

arclance · 2012-07-16 03:38:29

Actually that memory leak seems to be related to "idle_threshold" when using synergy.
It seems to happen when a message times out while the mouse cursor is on the other computer.
When the mouse returns to the computer running dunst the dunst window does not react to clicks and the memory leak starts.

I set "idle_threshold" to 0 to see what effect that has.

Last edited by arclance (2012-07-16 14:06:46)

thevdude · 2012-07-16 14:03:34

idle_threshold seemed a bit hacky last time I looked at the source.

Open an issue on github for it? https://github.com/knopwob/dunst

arclance · 2012-07-16 14:05:58

I was going to do that later today when I have time to write a good bug report.

It is working fine now that I have disabled it by setting "idle_threshold = 0", until it crashes on a Segmentation fault but I don't know what causes that.

I suppose I need to file two bug reports one for idle_threshold and one for the Segmentation fault.

I have a python script that gets the evolution email notifications from dbus and displays them in conky so the bugs are not a big problem for me right now but I would like them to be fixed so I can see notifications from other programs.

Last edited by arclance (2012-07-16 14:47:29)

arclance · 2012-07-16 17:42:16

I filed the bug reports one for "idle_threshold" and one for the Segmentation fault.

knopwob · 2012-07-24 12:48:17

Hi,

the memory-leak and segfaults reported by arclane should be resolved and I'll let the maintainer of the debian package know about the missing "provides" of notification-daemon.

Since my last post some contributors and I have added a few new features (those are not yet included in the debian package. I'll release version 0.3.0 soon, unless someone reports any major issues)

new option to align text left/center/right
new option to show age of notification
new keybinding to close all notifications
merge duplicate notifications ("(2) there are two duplicates of me")
if the geometry is set to a height of 1, indicate_hidden is set to true and there are hidden messages, the "(x more)" text is appended to the notification (e.g. no new line for that text)
The old way of defining keybindings has been depricated, but It should still work for the moment (unless I've broken it with the update).The new way has the advantage that you can have different sets of modifiers for different shortcuts.
Dunst can now detect which monitor has the focus and display its notifications on that monitor. You can define wether dunst should follow the mouse cursor, the keyboard focus or nothing.
To be able to follow the keyboard focus dunst needs the _NET_ACTIVE_WINDOW property to be set by the windowmanager. This should be the case with almost all modern windowmanagers.

More information about the new options can be found in the example dunstrc.

arclance · 2012-07-24 14:29:07

knopwob wrote:

Hi,
the memory-leak and segfaults reported by arclane should be resolved and I'll let the maintainer of the debian package know about the missing "provides" of notification-daemon.

I already let the maintainer know about the missing provides and he fixed it.

arclance · 2012-07-24 15:14:14

@ knopwob
Something in the new dunstrc just does not seem right to me.

    # redisplay last message(s)
    history = ctrl+grave

I don't think there is a grave key.

knopwob · 2012-07-24 15:36:59

arclance wrote:

@ knopwob
Something in the new dunstrc just does not seem right to me.
    # redisplay last message(s)
    history = ctrl+grave
I don't think there is a grave key.

On my US keyboard layout it's the one to write backticks (`) and with shift the tilde (~) above TAB and left of "1" .

arclance · 2012-07-24 15:43:13

knopwob wrote:

arclance wrote:
@ knopwob
Something in the new dunstrc just does not seem right to me.
    # redisplay last message(s)
    history = ctrl+grave
I don't think there is a grave key.
On my US keyboard layout it's the one to write backticks (`) and with shift the tilde (~) above TAB and left of "1" .

I have never heard of that key being called that before, just the "tilde" key.
You might want to add a comment about the key name to dunstrc if you find other people have not heard of it either.

Digit · 2012-11-23 04:22:30

mind if i join the dunst party? smile

unfortunately, it only seems to want to run as root just now (whichis relatively useless, since it only notifies of things run as root).

when attempting to run as user, i get:

Connection Error (Failed to connect to socket /tmp/dbus-9FtCm9WNzi: Connection refused)
dbus_con == NULL

dbus is something i never understood. is this some silly simple trivial matter to overcome? is it just a dbus issue? or does the fault lay with dunst, debian (i'm on waldorf), or just plain old pabkac somewhere along the line?

my googlings on the matter were unfruitful.

i utterly love the idea of dunst though. it should look great along with tickr, dmenu, i3status (or is it called i3bar?) (or scrotwm's bar or xmobar or whatever i have at the time), and whatever else in a similar ilk i get later on. ;D dunst... right up my street. feels closer to tickr-like now, rather than dwm/dmenu-like, with all the convenient config'ing n features being added.

+ amen for no qt dep. smile

arclance · 2012-11-23 04:40:29

^ There should be a dbus session for each user that is logged into your computer along with the one for root.
The users session-bus is started at login, I believe by your login manager.

If you already have another notification daemon running and it works you have a session bus for that user since all notifications are sent over dbus.

You can check if you have a session-bus after you are logged in by checking for a "dbus-daemon" process that includes "--session" in the command that was used to start it.
You should also have a "dbus-daemon --system" that is for root dbus messages.

If you have a session-bus I am not sure about that error since dunst works for me in Debian.

Last edited by arclance (2012-11-23 19:08:51)

CrunchBang Linux

SEARCH

#51 2012-07-16 03:00:09

Re: dunst - a dmenu-ish notification system

#52 2012-07-16 03:38:29

Re: dunst - a dmenu-ish notification system

#53 2012-07-16 14:03:34

Re: dunst - a dmenu-ish notification system

#54 2012-07-16 14:05:58

Re: dunst - a dmenu-ish notification system

#55 2012-07-16 17:42:16

Re: dunst - a dmenu-ish notification system

#56 2012-07-24 12:48:17

Re: dunst - a dmenu-ish notification system

#57 2012-07-24 14:29:07

Re: dunst - a dmenu-ish notification system

#58 2012-07-24 15:14:14

Re: dunst - a dmenu-ish notification system

#59 2012-07-24 15:36:59

Re: dunst - a dmenu-ish notification system

#60 2012-07-24 15:43:13

Re: dunst - a dmenu-ish notification system

#61 2012-11-23 04:22:30

Re: dunst - a dmenu-ish notification system

#62 2012-11-23 04:40:29

Re: dunst - a dmenu-ish notification system

Board footer