Hi,
>
>There is no procps in the ultra-minimal chroots used by piuparts :-)
>I have procps in my pbuilder chroot, too, but mainly because *my*
>metapackage depends on it. But I had no problems removing it (after my
>metapackage).
also in my pbuilder it is installed by default but I can remove it
>SERVER_PID=`ps -U \`whoami\` | grep VBoxSVC | awk '{ print $1 }'`
>if [ -z "$SERVER_PID" ]; then
> # Server not running yet/anymore, cleanup socket path.
> # See IPC_GetDefaultSocketPath()!
> if [ -n "$LOGNAME" ]; then
> rm -rf /tmp/.vbox-$LOGNAME-ipc > /dev/null 2>&1
> else
> rm -rf /tmp/.vbox-$USER-ipc > /dev/null 2>&1
> fi
>fi
yes, I reported upstream, and I added the runtime dependency on virtualbox,
because of that code,
but I can't push on alioth because it is down right now :)
>So it removes the socket path unconditionally because SERVER_PID is
>always empty due to missing ps. But if vbox is usually run as a normal
>user, there shouldn't be any VBoxSVC processes running as root, or am I
>mistaken here?
VBoxSVC is run as root, because the ext-pack during postinst is run as root.
quoting upstream
"the processes only "happen to run as root", because the package install is
done as root (and extpack install actually needs root privs to put files in the
right place)"
I think there is no bug here
>PS: looks like I cannot reproduce the leftover process problem any more
>while testing the latest version of virtualbox(-ext-pack) with piuparts
>(without doing any changes on my side)
well, strange then, because we didn't fix the issue. Maybe a newer virtualbox
version is faster in exit and doesn't spot the issue?
I'll upload a virtualbox with the added runtime dependency on the next upload
(I can't promise I'll do it soon, because I don't think the normal user can
have a graphical system without procps installed)
cheers,
(and thanks for the help!)
Gianfranco
