[systemd-devel] [PATCH v6 3/3] run: introduce timer support option

[Lennart Poettering]

On Tue, 09.12.14 00:03, WaLyong Cho (walyong.cho at samsung.com) wrote:

>          } else {
>                  log_error("Unknown assignment %s.", assignment);
>                  return -EINVAL;
> diff --git a/src/run/run.c b/src/run/run.c
> index 85eb052..e145784 100644
> --- a/src/run/run.c
> +++ b/src/run/run.c
> @@ -30,6 +30,7 @@
>  #include "env-util.h"
>  #include "path-util.h"
>  #include "bus-error.h"
> +#include "calendarspec.h"
>  
>  static bool arg_scope = false;
>  static bool arg_remain_after_exit = false;
> @@ -47,27 +48,45 @@ static int arg_nice = 0;
>  static bool arg_nice_set = false;
>  static char **arg_environment = NULL;
>  static char **arg_property = NULL;
> +static bool with_timer = false;

Hmm, maybe we should turn this boolean into a function? I mean, static
variables are something we should keep at a minimum. We do them for
command line arguments, since they are kinda global anyway, but I
think we should otherwise avoid them, especially if they are
effectively redundant anyway.

Hence, my suggestions would be to turn this into a call:

        static bool with_timer(void) {
                return arg_on_active || arg_on_boot || arg_on_startup || arg_on_unit_active || arg_on_unit_inactive || arg_on_calendar;
        }

Or so?

> +static int start_transient_service(
> +                sd_bus *bus,
> +                char **argv,
> +                sd_bus_error *error) {
> +
> +        _cleanup_bus_message_unref_ sd_bus_message *m = NULL;
> +        _cleanup_free_ char *service = NULL;
> +        char *state = NULL;
> +        int r;
> +
> +        assert(bus);
> +        assert(argv);
> +
> +        if (arg_unit) {
> +                service = unit_name_mangle_with_suffix(arg_unit, MANGLE_NOGLOB, ".service");
> +                if (!service)
> +                        return log_oom();
> +
> +                if (get_unit_state_by_name(bus, service, &state) == 0) {
> +                        log_error("Unit %s is already exist with %s state.", service, state);
> +                        return -EEXIST;
> +                }

What's the rationale for checking the unit state here? I mean, PID 1
should fail anyway in this case, do we realy need to check this name
in advance? I mean, such a check is necessarily racy, since the unit
might appear while between our client-side check and the actual
execution of the service, no?

Lennart

-- 
Lennart Poettering, Red Hat