Re: [PATCH 01/21] perf tools: Introduce struct add_entry_iter

[Namhyung Kim <namhyung@kernel.org>]

On Sun, 5 Jan 2014 16:55:10 +0100, Jiri Olsa wrote:
> On Tue, Dec 24, 2013 at 05:22:07PM +0900, Namhyung Kim wrote:
>> From: Namhyung Kim <namhyung.kim@lge.com>
>
> SNIP
>
>> +
>> +out:
>> +	iter->curr++;
>> +	return err;
>> +}
>> +
>> +static int
>> +iter_finish_branch_entry(struct add_entry_iter *iter,
>> +			 struct addr_location *al __maybe_unused)
>> +{
>> +	free(iter->priv);
>> +	iter->priv = NULL;
>
> so branch_info is duplicated in the hist_entry__new.. it's not
> easy to find why this one gets freed while mem_info stays ;-)

The branch_info is an array and each entry gets its own part in the
array - so entries alloc & copy the (part of) branch info as the branch
info iterated.  But mem info is a single data for a single entry so a
new entry steals ownership of the mem info when created and not freed.

Maybe we can alloc & copy another mem info when adds an entry and free
the original at iter->finish_entry(), but I felt it's unnecessary..

>
> some comment about that here would help

How about this?

  /*
   * The mem_info was either already freed in add_hist_entry() or
   * passed to a new hist entry by hist_entry__new().
   */


Thanks,
Namhyung