Re: [PATCH v4 8/8] trace: Add alternative synthetic event trace action syntax

[Tom Zanussi <zanussi@kernel.org>]

Hi Masami,

On Wed, 2018-09-19 at 08:54 +0900, Masami Hiramatsu wrote:
> Hi Tom,
> 
> On Tue, 18 Sep 2018 14:16:43 -0500
> Tom Zanussi <zanussi@kernel.org> wrote:
> 
> > Hi Masami,
> > 
> > On Wed, 2018-09-19 at 03:54 +0900, Masami Hiramatsu wrote:
> > > Hi Tom,
> > > 
> > > On Mon, 10 Sep 2018 14:10:46 -0500
> > > Tom Zanussi <zanussi@kernel.org> wrote:
> > > 
> > > > From: Tom Zanussi <tom.zanussi@linux.intel.com>
> > > > 
> > > > Add a 'trace(synthetic_event_name, params)' alternative to
> > > > synthetic_event_name(params).
> > > > 
> > > > Currently, the syntax used for generating synthetic events is
> > > > to
> > > > invoke synthetic_event_name(params) i.e. use the synthetic
> > > > event
> > > > name
> > > > as a function call.
> > > > 
> > > > Users requested a new form that more explicitly shows that the
> > > > synthetic event is in effect being traced.  In this version, a
> > > > new
> > > > 'trace()' keyword is used, and the synthetic event name is
> > > > passed
> > > > in
> > > > as the first argument.
> > > 
> > > Hmm, what is the advantage of adding this new form?
> > > 
> > 
> > There's no real advantage other than user preference - Namhyung
> > thought
> > that since the event-name-as-function-call actions are all defined
> > as
> > ACTION_TRACE, there should also be an explicit 'trace' action.
> 
> Ah, got it. Would this needs documentation and testcase update too?
> 

The first part of the patch does contain the Documentation change to
trace/histograms.txt.

But I didn't do a testcase for it, will do that in the next iteration.

> > 
> > So I added it as alternative syntax - the event-name-as-function-
> > call
> > form remains unchanged.
> > 
> > By the way, I also have a patch implementing your alternative
> > syntax
> > change, where if you have only one handler, you can do away with
> > the
> > explicit action.handler form e.g.
> > 
> >   # echo 'hist:keys=next_pid:wakeup_lat=common_timestamp.usecs-ts0: 
> > \
> >     onmax($wakeup_lat): \
> >     save(next_prio,next_comm,prev_pid,prev_prio,prev_comm):snapshot
> > () \
> 
> Hmm, in this case, I think comma-connected syntax will be clearer
> when
> the action is kicked.
> 
>      onmax($wakeup_lat).save(next_prio,next_comm,prev_pid,prev_prio,p
> rev_comm),snapshot()
> 
> any thought?
> 

With the above I was trying to implement something close to your
original suggestion of:

<actions> onchange(<var>)

So what I described was:

<actions>:onchange(<var>)

where <actions> would expand to action1[:action2:..]

But as I mentioned both only make sense if you only have one
onchange(<var>).

Your new version:

onchange(<var>).action1[,action2,..]

is probably better as it would allow for multiple onchange<var> and/or
other handlers.

Thanks,

Tom