Re: perf: definition of a "regression"

[Vince Weaver]

On Sat, 28 May 2011, Peter Zijlstra wrote:

> On Fri, 2011-05-27 at 23:38 -0400, Vince Weaver wrote:
> > on that note (and while trying to document exactly what the ioctls do) it 
> > seems that a PERF_EVENT_IOC_REFRESH with an argument of anything higher 
> > than one does not work on kernels 2.6.36 and newer.  The behavior acts
> > as if 1 was passed, even if you pass in, say, 3.
> 
> Urgh, no that should definitely work. Thanks for the test-case, I'll
> work on that (probably not until Monday though, but who knows).

So wait, the two regressions I found in 2.6.37 are WONTFIX because they 
are too old, even though they break existing userspace code?

And this older regression in 2.6.36 is going to be fixed, even though
perf, PAPI, and libpfm4 don't trigger the buggy functionality at all?

I think it's time to redefine the PERF_EVENT_IOC_REFRESH ioctl to just
refresh once (as that's what it actually does on 2.6.36 - 2.6.39) and
if we need to refresh multiple we should add a new 
PERF_EVENT_IOC_REFRESH_COUNT ioctl.

I know I am being difficult, but the perf-event ABI is a mess to program 
for in a backward compatible fashion.

Vince
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/