Re: [PATCH] perf trace: Check if MAP_32BIT is defined (again)

[Jiri Olsa]

On Fri, Dec 16, 2016 at 11:02:47AM -0300, Arnaldo Carvalho de Melo wrote:
> Em Thu, Dec 15, 2016 at 08:56:54PM +0100, Jiri Olsa escreveu:
> > There might be systems where MAP_32BIT is not defined,
> > like some some RHEL7 powerpc versions.
> > 
> > Signed-off-by: Jiri Olsa <jolsa@xxxxxxxxxx>
> > Cc: Kyle McMartin <kyle@xxxxxxxxxx>
> > Cc: David Ahern <dsahern@xxxxxxxxx>
> > Cc: Namhyung Kim <namhyung@xxxxxxxxxx>
> > Cc: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>
> > Fixes: 277cf08f3feb ("perf trace beauty mmap: Fix defines for non !x86_64")
> 
> Ouch, so we need what was removed in:
> 
> Fixes: 256763b01741 ("perf trace beauty mmap: Add more conditional defines")
> 
> I.e. the ifdef at the switch case entry, not the removal of the
> conditional definition expecting that it would be available in the arch
> header, as in some cases those are not there. So I'm replacing the Fixme
> cset, ok?

sure, ok

> 
> Thanks for taking the time to fill in that tag, appreciated!
> 
> Also this all needs a revamp to support cross 'perf trace'ing, i.e.
> collecting a file on arch A and doing 'perf trace -i perf.data' on a
> different machine with a CPU fo a different arch, where either per arch
> beautifiers are needed or a dynamic switch, i.e. the value for some
> define is different between archs....
> 
> Patch applied, thanks,

thanks,
jirka