Re: [PATCH] libvirt-guests.sh: fix log output with new systemd

On Thu, Apr 30, 2020 at 2:10 PM Michal Privoznik <mprivozn(a)redhat.com&gt; wrote: > > On 4/30/20 11:59 AM, Christian Ehrhardt wrote: > > Newer systemd is too smart, it detects the PIDs of the gettext calls > > and due to that a log lists libvirt-guests.sh with many different PIDs. > > > > Furthermore it struggles to collect the output in time, so it can be > > truncated and overall looks like: > > libvirt-guests.sh[37986]: Running guests on default URI: > > libvirt-guests.sh[37995]: Running guests on > > libvirt-guests.sh[37977]: R > > > > Gather the gettext result into a local variable and printing > > the value from libvirt-guests.sh itself fixes both issues. > > > > Fixes: https://bugs.launchpad.net/ubuntu/+source/libvirt/+bug/1875708 > > > > Signed-off-by: Christian Ehrhardt <christian.ehrhardt(a)canonical.com&gt; > > --- > > tools/libvirt-guests.sh.in | 74 +++++++++++++++++++++----------------- > > 1 file changed, 42 insertions(+), 32 deletions(-) > > > > diff --git a/tools/libvirt-guests.sh.in b/tools/libvirt-guests.sh.in > > index 7af24dab3b..324abe3623 100644 > > --- a/tools/libvirt-guests.sh.in > > +++ b/tools/libvirt-guests.sh.in > > @@ -28,6 +28,16 @@ test ! -r "$sysconfdir"/rc.d/init.d/functions || > > # Make sure this file is recognized as having translations: _("dummy") > > . &quot;@bindir(a)&quot;/gettext.sh > > > > +# Avoid output being listed under gettext PID and being truncated > > +unbuffered_gettext () { > > + msg="$(gettext "$1")" > > + echo "$msg" > > +} > > +unbuffered_eval_gettext () { > > + msg="$(eval_gettext "$1")" > > + echo "$msg" > > +} > > + > > Ah, is this the case where gettext prints the text in many sequences > whereas echo does it all at once? Or what do you think is happening > here? I don't quite understand how buffering steps into this. You have two effects: - systemd tries to detect and report which PID the output is from when something it printed. Therefore it appears to a user as if the script itself would have been invoked multiple times while actually is will be the PIDs of the different gettext calls - I assume that to get this content it is hitchhiking the output streams somehow - the gettext processes are very very short lived and it seems when they are exiting some output of them might be lost.