On Fri, Jun 14, 2019 at 12:10:48PM +0200, Philippe Mathieu-Daudé wrote:
> On 6/14/19 11:33 AM, Stefan Hajnoczi wrote:
> > On Fri, Jun 14, 2019 at 10:13:04AM +1000, Alexey Kardashevskiy wrote:
> >>
> >>
> >> On 13/06/2019 23:08, Philippe Mathieu-Daudé wrote:
> >>> Hi Alexey,
> >>>
> >>> On 6/13/19 7:09 AM, Alexey Kardashevskiy wrote:
> >>>> This adds a trace point which prints every loaded image. This includes
> >>>> bios/firmware/kernel/initradmdisk/pcirom.
> >>>>
> >>>> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> >>>> ---
> >>>>
> >>>> The example for a pseries guest:
> >>>>
> >>>> loader_write_rom slof.bin: @0x0 size=0xe22e0 ROM=0
> >>>> loader_write_rom phdr #0: /home/aik/t/vml4120le: @0x400000 size=0x13df000 ROM=0
> >>>> loader_write_rom /home/aik/t/le.cpio: @0x1ad0000 size=0x9463a00 ROM=0
> >>>
> >>> I find the "ROM=0" part confuse, maybe you can change to "ROM:false".
> >>
> >> How? I mean I can do that in the code as rom->isrom?"true":"false" and
> >> make trace point accept "%s" but it is quite ugly and others seem to
> >> just use %d for bool.
> > 
> > Yes, %d is the convention for bool.  Perhaps you can name it "is_rom"
> > instead of "ROM".  That way the name communicates that this is a boolean
> > value.
> 
> I agree a boolean is easier to parse for trace analyzer tools than a
> "true/false" string.
> 
> Stefan, is it possible to add a boolean format string to the backends?
> For example the 'log' backend would log it as "true"/"false".

Yes.  The format string in trace-events is not used by all trace
backends, and they are free to do whatever they like (e.g. by looking at
the C types of the trace event arguments).

Stefan