On Thu, 29 Aug 2019 14:39:43 +0300 Tzvetomir Stoyanov <tz.stoyanov@xxxxxxxxx> wrote: > On Wed, Aug 28, 2019 at 10:59 PM Steven Rostedt <rostedt@xxxxxxxxxxx> wrote: > > > > On Wed, 14 Aug 2019 11:47:04 +0300 > > "Tzvetomir Stoyanov (VMware)" <tz.stoyanov@xxxxxxxxx> wrote: > > > > > void tracecmd_set_quiet(int quiet); > > > int tracecmd_get_quiet(void); > > > > I rather make this a parameter to the descriptor: > > > > tracecmd_set_quiet(struct tracecmd_output *handle, bool quiet); > > tracecmd_get_quiet(struct tracecmd output *handle); > > > > As we may have multiple handles and perhaps we don't want all of them > > quiet. > > > It makes sense to have "quiet" per tracecmd_output handler, but when I > started to modify the code, > I noticed one flow where tracecmd_get_quiet() is used and there is no > tracecmd_output handler available - > in check_plugin(). We have either to drop the usage of > tracecmd_get_quiet() in check_plugin(), or leave the scope of > "quiet" to be for the whole library. We can still have a global variable in trace-record.c called quiet. And that gets set by the parameter: case OPT_quite: case 'q': quiet = 1; break; [..] tracecmd_set_quiet(handle, quiet); This is the proper way to handle it. The functions local to trace-record.c, can just use its own quiet state variable, but when we set quiet, we use the tracecmd_set_quiet() to notify the library that it too should be quiet. -- Steve
|