Re: [PATCH 22/27] perf tools: Enable --children option by default

[Namhyung Kim]

Hi Andres,

On Tue, 29 Jul 2014 13:45:48 +0200, Andres Freund wrote:
> On 2014-07-29 15:52:46 +0900, Namhyung Kim wrote:
>> On Mon, 28 Jul 2014 21:23:26 +0200, Andres Freund wrote:
>> > On 2014-05-29 12:58:25 +0900, Namhyung Kim wrote:
>> >> Now perf top and perf report will show children column by default if
>> >> it has callchain information.
>> >> 
>> >> Requested-by: Ingo Molnar <mingo@xxxxxxxxxx>
>> >> Tested-by: Rodrigo Campos <rodrigo@xxxxxxxxxxx>
>> >> Tested-by: Arun Sharma <asharma@xxxxxx>
>> >> Acked-by: Jiri Olsa <jolsa@xxxxxxxxxx>
>> >> Cc: Frederic Weisbecker <fweisbec@xxxxxxxxx>
>> >> Signed-off-by: Namhyung Kim <namhyung@xxxxxxxxxx>
>> >
>> > Stupid question: Why? This breaks user's usage pattern in a rather
>> > significant way because the new output is pretty much entirely useless
>> > in many cases, especially for userland programs. I think the new option
>> > is great, but why is it suddenly the default?
>> 
>> Yes, I knew it'd annoy some existing users like you :)  But I hoped it'd
>> be useful for new and (at least, some of) old users to see children/
>> cumulative overheads by default.  If it lies under an option, it
>> wouldn't have a chance to be used by normal users who simply run perf
>> record and report most of time.
>
> By that argument every new feature would need to be enabled by
> default. Which normally isn't what happens.

Yes, I know and that's what we usually do.  What I wanted to say was
it's a special case. :)


>
>> If you think it's entirely useless for you cases, you can set it to off
>> using ~/.perfconfig file:
>> 
>>   [report]
>>   children = false
>
> Already done that ;).

Good. :)


>
> The problem is still that I can't use --no-children on older systems, so
> I can't write benchmarking scripts anymore that don't take depend on the
> perf version...

Hmm.. I don't understand what's your problem.  If you set the config
option you don't need to use --no-children at all and it'd work in a
same way for old and new versions, no?


>
> I guess it's too late to argue effectively though :)

Well, I think it can be changed later, if enough number of users yell at
me - now I have two. :)

Thanks,
Namhyung
--
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/