Re: [PATCH] tracing: fix TRACE_EVENT power tracepoint creation

[Mathieu Desnoyers <mathieu.desnoyers@efficios.com>]

* Steven Rostedt (rostedt@goodmis.org) wrote:
> On Thu, 2010-10-28 at 07:03 -0400, Mathieu Desnoyers wrote:
> 
> > Given that trace/power.h is inluded in more than one .c file, generating
> > DEFINE_TRACE() when CREATE_TRACE_POINTS is not set does not work. So what
> > alternative do you suggest ?
> 
> Right, DEFINE_TRACE should only be set when CREATE_TRACE_POINTS is set,
> and that is only set in the C file that the DEFINE_TRACE is to be
> defined.
> 
> In fact, I don't even see what bug you are trying to fix?
> 
> The file include/trace/define_trace.h converts all the TRACE_EVENTS()
> into DEFINE_TRACE() when CREATE_TRACE_POINTS is set. No other C file
> that includes power.h should have CREATE_TRACE_POINTS set. I guess if it
> did, the CONFIG_EVENT_TRACING would break too.
> 
> It should just work.

This is the culprit:

kernel/trace/ is not built when CONFIG_TRACING=n
kernel/trace/power-traces.o is only built when CONFIG_EVENT_TRACING=y

So the DEFINE_TRACE is never done when CONFIG_EVENT_TRACING=n or
CONFIG_TRACING=n.

What you are telling me is that changing the makefiles should be enough, and I
don't need to add those DEFINE_TRACE(), correct ?

Thanks,

Mathieu

-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com