Re: [PATCH] tracing: fix a build error on alpha and m68k

[Ingo Molnar <mingo@elte.hu>]

* Frédéric Weisbecker <fweisbec@gmail.com> wrote:

> But they should be able to use the tracing for several other tracers. 
> One other problem is that you need CONFIG_TRACE_IRQFLAGS (is that a 
> lockdep or just trace feature?) to use any tracer since the 
> tracing_cpumask file is always created for all tracer. This file has 
> tracing_cpumask_write() as a write operation, and this func uses 
> raw_local_funcs....
> 
> Perhaps we should disable the tracing_cpu_mask related things if 
> TRACE_IRQFLAGS in not configured?

to answer the "is that a lockdep or just trace feature" question:

trace-irqflags was first written by me for the (crude) ftrace-precursor 
latency tracer code in -rt, years ago. Then i reused it (and changed it 
alot) for upstream lockdep, two years ago. Then ftrace came in this year 
and reused it.

so it's rather symbiotic ;-)

So ... the tracers that rely on irqflags-tracing should definitely be 
limited to architectures that provide TRACE_IRQFLAGS. The core trace.c 
itself should probably not be restricted ... (and it should definitely 
build)

	Ingo