dbus user existance checking.

[Havoc Pennington]

Hi,

Thanks for the patch, some comments:

 1. I think we want a syslog=yes|no option in the config file, and 
    default yes for system daemon and no for session; my experience 
    from gconf is that admins aren't big fans of anything in syslog
    unless said thing requires their intervention.

    The obvious start to this fix is to have the field in the config
    file, and set a bool in BusConfigParser. It should work like 
    the BusLimits stuff (i.e. handle included files, as in 
    the patch Jon just posted).

    The problem after that is how to get the setting for each 
    BusContext to _dbus_warn and it probably means adding 
    a _dbus_warn_with_log(dbus_bool_t log, const char *format, )
    function or something like that then fixing up all the 
    _dbus_warn to honor the flag in the BusContext, if that 
    makes sense.

 2. So further along those lines, we sort of need to grep for 
    _dbus_warn and see how many of them would be irritating if 
    they ended up in syslog

 3. The configure.in check isn't quite right; the line:
    AC_CHECK_HEADERS(execinfo.h, [AC_CHECK_FUNCS(backtrace)])
    is checking for the execinfo.h header, and if that's found 
    checking for the backtrace function. This is then used 
    in dbus-sysdeps.c:
      #if defined (HAVE_BACKTRACE)

    So the vsyslog check has to be separate from the backtrace 
    check, and it's pointless unless we use HAVE_VSYSLOG in the 
    code somewhere.

Havoc