heftig
rtkit-daemon: Don't log debug messages by default
The rtkit-daemon service is a lot more verbose than other services when it doesn't have anything to do. Stop logging the debug messages by default to avoid flooding the system log.
This addresses issue #22.
I know this is the 3rd pull request to address the same issue. However, as the previous two proposals were never merged, I thought it would make sense to offer an alternative which may look more appealing.
The first proposal by KogasaPls in https://github.com/heftig/rtkit/pull/26 opted to leave the default behavior unchanged, and added a flag which you need to enable to actually get rid of the flood of debug messages. While I understand how being conservative can increase the chances of getting such a change accepted, it also has the drawback that users and package maintainers then have extra work to actually get the benefits of the change. Also, the review of the changes revealed possible problems which do not appear to have been addressed.
The second proposal by TriMoon in https://github.com/heftig/rtkit/pull/31 opted to filter out the flood of debug messages at the systemd level. While this does work, it has two drawbacks. First, it depends on systemd, so people using rtkit-daemon without systemd are left without a solution. Second, it does not prevent the messages from being generated in the first place, so from a performance perspective, this is an inferior solution, which looks more like wallpapering over the issue that actually addressing it.
My proposal is to change the default behavior to align rtkit-daemon with what all other daemons out there do, that is, don't print debug messages by default. This addresses all the concerns underlined above. The commit itself is also simple and self-contained, so the risk of regression is fairly low.
Is there any reason for the hesitation in merging this request? As someone who is only really an end-user of rtkit, it is slightly frustrating trying to debug system errors when rtkit decides it needs to send messages every X seconds that don't really seem to have much information in them. While I understand that it is possible to filter this out at the systemd level by using a drop-in file, a fix from upstream is obviously preferred.
