On Mon, 16 Mar 2020 at 11:00:09 +0100, Harald Dunkel wrote: > On 2020-03-16 09:20, Simon McVittie wrote: > > Do you use chromium, Chrome, Electron or qtwebengine? > > I am just the messenger, but AFAICT its Google Chrome. However, the process > id 3434 (mentioned in syslog) belongs to gnome-shell.
I'm fairly sure that's because Chrome has inherited a logging file descriptor from GNOME Shell. The PID is recorded at the time the file descriptor is opened, not at the time the message is actually logged. > I think its unwise to > generate appr 12000 syslog messages per minute, so I still think thats a bug > in gnome-shell. I don't think it is gnome-shell that is generating these messages. If you run Chrome from a terminal (gnome-terminal or xterm or similar), or ask the user to do so if you are reporting this bug on behalf of someone else, then I suspect you will find that the warnings stop appearing in the system log, and start appearing in the terminal instead. Another way to check whether I'm correct would be to edit Chrome's .desktop file(s) and change all the Exec= lines (there might be more than one) to use systemd-cat, something like this: [Desktop Entry] ... -Exec=/path/to/chrome-binary %U +Exec=systemd-cat -t chrome /path/to/chrome-binary %U > > How do you launch the chromium-based app? > > Parent process of Chrome is gnome-shell, package version 3.30.2-11~deb10u1. That's consistent with my theory, but doesn't fully answer my question. There are several ways to launch an app from GNOME Shell, which end up in different code paths, including: - open overview ("Activities"), click on an app - open overview, click-and-hold or right-click on an app, "New Window" Some Shell extensions might also influence how apps are launched. Are you (or the user, if you are passing on a bug report from someone else) using any Shell extensions? If so, which ones? I think the current situation is that just clicking on an app will usually result in its stdout/stderr being logged as though they came from that app, but "New Window" will often result in its stdout/stderr being logged as though they came from GNOME Shell. I've opened an upstream issue about that, but it will not be straightforward to fix (it's likely to require new API in GLib). I don't know whether that's the case you're hitting here, though. smcv