[Pkg-ayatana-devel] Bug#783125: dbus-daemon -- produce a log of syslog messages

[Simon McVittie]

Control: reassign 783125 indicator-applet
Control: retitle 783125 many GVariant critical warnings logged to syslog

On 22/04/15 18:24, Jörg Frings-Fürst wrote:
> Apr 22 18:40:52 merkur com.canonical.indicator.messages[4248]: (process:4574):
> GLib-CRITICAL **: g_variant_get_string: assertion 'g_variant_is_of_type (value,
> G_VARIANT_TYPE_STRING) || g_variant_is_of_type (value,
> G_VARIANT_TYPE_OBJECT_PATH) || g_variant_is_of_type (value,
> G_VARIANT_TYPE_SIGNATURE)' failed

That message is actually from whatever your implementation of
com.canonical.indicator.messages is; I think that means indicator-applet
(but please reassign elsewhere if not). If you look in
/usr/share/dbus-1/services for a file that mentions
Name=com.canonical.indicator.messages, these messages are from whatever
is in the Exec= line in that file.

> "ps -Af | grep 4248" has the output:
> 
> jff       4248     1  0 06:25 ?        00:00:41 /usr/bin/dbus-daemon --fork
> --print-pid 5 --print-address 7 --session

The process ID is misleading, unfortunately. dbus-daemon changes the
process name that is reported to systemd when it starts a session
service (so that at least the messages from session services aren't all
blamed on "dbus-daemon[4248]" which would be even more misleading), but
it isn't able to change the process ID that is reported to the Journal,
because the APIs that it uses to redirect stdout/stderr to the Journal
don't seem to be async-signal-safe, which means they could cause a crash
if called after fork() in the child process that already has the new pid.

So this isn't ideal, but it's the least-bad thing that dbus-daemon can
do in the circumstances.

    S