[PATCH] perf tools: Fix "Command" sort_entry's cmp and collapse function

[PATCH] perf tools: Fix "Command" sort_entry's cmp and collapse function

[Jiri Olsa]

Currently the se_cmp and se_collapse use pointer comparison,
which is ok for for testing equality of strings. It's not ok
as comparing function for rbtree insertion, because it gives
different results based on current pointer values.

We saw test 32 (hists cumulation test) failing based on different
environment setup. Having all sort functions straightened fix the
test for us.

Reported-by: Jan Stancek <jstancek@redhat.com>
Link: http://lkml.kernel.org/n/tip-tklp6y27bseqjibcwn0py9up@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/util/sort.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
index 4593f36ecc4c..09d4696fd9a1 100644
--- a/tools/perf/util/sort.c
+++ b/tools/perf/util/sort.c
@@ -89,14 +89,14 @@ static int64_t
 sort__comm_cmp(struct hist_entry *left, struct hist_entry *right)
 {
 	/* Compare the addr that should be unique among comm */
-	return comm__str(right->comm) - comm__str(left->comm);
+	return strcmp(comm__str(right->comm), comm__str(left->comm));
 }
 
 static int64_t
 sort__comm_collapse(struct hist_entry *left, struct hist_entry *right)
 {
 	/* Compare the addr that should be unique among comm */
-	return comm__str(right->comm) - comm__str(left->comm);
+	return strcmp(comm__str(right->comm), comm__str(left->comm));
 }
 
 static int64_t
-- 
1.9.3

Re: [PATCH] perf tools: Fix "Command" sort_entry's cmp and collapse function

[Namhyung Kim]

Hi Jiri,

CC-ing Frederic as he wrote the comm change.

On Fri, May 15, 2015 at 05:54:28PM +0200, Jiri Olsa wrote:
> Currently the se_cmp and se_collapse use pointer comparison,
> which is ok for for testing equality of strings. It's not ok
> as comparing function for rbtree insertion, because it gives
> different results based on current pointer values.
> 
> We saw test 32 (hists cumulation test) failing based on different
> environment setup. Having all sort functions straightened fix the
> test for us.

Can you elaborate it?

AFAIK comm string is shared among threads so pointer comparison and
'strcmp == 0' should have same result..

Thanks,
Namhyung


> 
> Reported-by: Jan Stancek <jstancek@redhat.com>
> Link: http://lkml.kernel.org/n/tip-tklp6y27bseqjibcwn0py9up@git.kernel.org
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  tools/perf/util/sort.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
> index 4593f36ecc4c..09d4696fd9a1 100644
> --- a/tools/perf/util/sort.c
> +++ b/tools/perf/util/sort.c
> @@ -89,14 +89,14 @@ static int64_t
>  sort__comm_cmp(struct hist_entry *left, struct hist_entry *right)
>  {
>  	/* Compare the addr that should be unique among comm */
> -	return comm__str(right->comm) - comm__str(left->comm);
> +	return strcmp(comm__str(right->comm), comm__str(left->comm));
>  }
>  
>  static int64_t
>  sort__comm_collapse(struct hist_entry *left, struct hist_entry *right)
>  {
>  	/* Compare the addr that should be unique among comm */
> -	return comm__str(right->comm) - comm__str(left->comm);
> +	return strcmp(comm__str(right->comm), comm__str(left->comm));
>  }
>  
>  static int64_t
> -- 
> 1.9.3
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Re: [PATCH] perf tools: Fix "Command" sort_entry's cmp and collapse function

[Jiri Olsa]

On Mon, May 18, 2015 at 10:38:50AM +0900, Namhyung Kim wrote:
> Hi Jiri,
> 
> CC-ing Frederic as he wrote the comm change.
> 
> On Fri, May 15, 2015 at 05:54:28PM +0200, Jiri Olsa wrote:
> > Currently the se_cmp and se_collapse use pointer comparison,
> > which is ok for for testing equality of strings. It's not ok
> > as comparing function for rbtree insertion, because it gives
> > different results based on current pointer values.
> > 
> > We saw test 32 (hists cumulation test) failing based on different
> > environment setup. Having all sort functions straightened fix the
> > test for us.
> 
> Can you elaborate it?
> 
> AFAIK comm string is shared among threads so pointer comparison and
> 'strcmp == 0' should have same result..

well pointers (A, B) substraction will get either negative (A < B) or
possitive number (A > B). Now if your environment happens to have those
pointers switched you'll get different result for same strings.

Following python script forced the 'perf test 32' to switch those pointers
and screwed the Command sorting.. it's not very reliable on upstream for
some reason.. but very reproducible on RHEL7

---
#!/bin/python

import os

env={}

env['PATH'] = "/tmp:"*500

os.execve("./perf", ["perf", "test", "32", "-v"], env)
---

jirka