[PATCH] perf symbols: reuse of name in dso__load when starting second pass

[PATCH] perf symbols: reuse of name in dso__load when starting second pass

[David Ahern]

continue statement in default case applies when a second pass is
wanted as well. As it stands the code drops down to the open
re-using the value in name from the previous origin attempt.

Signed-off-by: David Ahern <daahern@cisco.com>
---
 tools/perf/util/symbol.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
index d30136e..861be8b 100644
--- a/tools/perf/util/symbol.c
+++ b/tools/perf/util/symbol.c
@@ -1488,8 +1488,8 @@ int dso__load(struct dso *self, struct map *map, symbol_filter_t filter)
 			if (want_symtab) {
 				want_symtab = 0;
 				self->origin = DSO__ORIG_BUILD_ID_CACHE;
-			} else
-				continue;
+			}
+			continue;
 		}
 
 		/* Name is now the name of the next image to try */
-- 
1.7.2.3

Re: [PATCH] perf symbols: reuse of name in dso__load when starting second pass

[Arnaldo Carvalho de Melo]

Em Sun, Nov 21, 2010 at 08:03:50AM -0700, David Ahern escreveu:
> continue statement in default case applies when a second pass is
> wanted as well. As it stands the code drops down to the open
> re-using the value in name from the previous origin attempt.
> 
> Signed-off-by: David Ahern <daahern@cisco.com>
> ---
>  tools/perf/util/symbol.c |    4 ++--
>  1 files changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
> index d30136e..861be8b 100644
> --- a/tools/perf/util/symbol.c
> +++ b/tools/perf/util/symbol.c
> @@ -1488,8 +1488,8 @@ int dso__load(struct dso *self, struct map *map, symbol_filter_t filter)
>  			if (want_symtab) {
>  				want_symtab = 0;
>  				self->origin = DSO__ORIG_BUILD_ID_CACHE;
> -			} else
> -				continue;
> +			}
> +			continue;
>  		}

But we want to restart at DSO__ORIG_BUILD_ID_CACHE, right? If you do
your change we keep a existing mistake and will instead restart at
DSO__ORIG_BUILD_ID_CACHE + 1, please take a look to validate my
understanding that we need to set self->origin to
DSO__ORIG_BUILD_ID_CACHE - 1.

- Arnaldo

Re: [PATCH] perf symbols: reuse of name in dso__load when starting second pass

[David S. Ahern]

On 11/21/10 15:03, Arnaldo Carvalho de Melo wrote:
> Em Sun, Nov 21, 2010 at 08:03:50AM -0700, David Ahern escreveu:
>> continue statement in default case applies when a second pass is
>> wanted as well. As it stands the code drops down to the open
>> re-using the value in name from the previous origin attempt.
>>
>> Signed-off-by: David Ahern <daahern@cisco.com>
>> ---
>>  tools/perf/util/symbol.c |    4 ++--
>>  1 files changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
>> index d30136e..861be8b 100644
>> --- a/tools/perf/util/symbol.c
>> +++ b/tools/perf/util/symbol.c
>> @@ -1488,8 +1488,8 @@ int dso__load(struct dso *self, struct map *map, symbol_filter_t filter)
>>  			if (want_symtab) {
>>  				want_symtab = 0;
>>  				self->origin = DSO__ORIG_BUILD_ID_CACHE;
>> -			} else
>> -				continue;
>> +			}
>> +			continue;
>>  		}
> 
> But we want to restart at DSO__ORIG_BUILD_ID_CACHE, right? If you do
> your change we keep a existing mistake and will instead restart at
> DSO__ORIG_BUILD_ID_CACHE + 1, please take a look to validate my
> understanding that we need to set self->origin to
> DSO__ORIG_BUILD_ID_CACHE - 1.

Correct. In working on the rootfs change I noticed that the default case
falls through to the open using the 'name' value set from the previous
origin so it is tried twice. This patch fixes that bug.

I also noted that the BUILD_ID_CACHE was skipped, but I was not sure if
that was desired or not. If it is then correcting the proper 'next
origin' is another trivial bug fix.

Some projects like single focused bug fixes for bisecting. I am fine
with combining if that is wanted.

David


> 
> - Arnaldo

Re: [PATCH] perf symbols: reuse of name in dso__load when starting second pass

[Arnaldo Carvalho de Melo]

Em Sun, Nov 21, 2010 at 05:50:46PM -0700, David S. Ahern escreveu:
> On 11/21/10 15:03, Arnaldo Carvalho de Melo wrote:
> > Em Sun, Nov 21, 2010 at 08:03:50AM -0700, David Ahern escreveu:
> >> continue statement in default case applies when a second pass is
> >> wanted as well. As it stands the code drops down to the open
> >> re-using the value in name from the previous origin attempt.
> >>
> >> Signed-off-by: David Ahern <daahern@cisco.com>
> >> ---
> >>  tools/perf/util/symbol.c |    4 ++--
> >>  1 files changed, 2 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
> >> index d30136e..861be8b 100644
> >> --- a/tools/perf/util/symbol.c
> >> +++ b/tools/perf/util/symbol.c
> >> @@ -1488,8 +1488,8 @@ int dso__load(struct dso *self, struct map *map, symbol_filter_t filter)
> >>  			if (want_symtab) {
> >>  				want_symtab = 0;
> >>  				self->origin = DSO__ORIG_BUILD_ID_CACHE;
> >> -			} else
> >> -				continue;
> >> +			}
> >> +			continue;

> > But we want to restart at DSO__ORIG_BUILD_ID_CACHE, right? If you do
> > your change we keep a existing mistake and will instead restart at
> > DSO__ORIG_BUILD_ID_CACHE + 1, please take a look to validate my
> > understanding that we need to set self->origin to
> > DSO__ORIG_BUILD_ID_CACHE - 1.
> 
> Correct. In working on the rootfs change I noticed that the default case
> falls through to the open using the 'name' value set from the previous
> origin so it is tried twice. This patch fixes that bug.
> 
> I also noted that the BUILD_ID_CACHE was skipped, but I was not sure if
> that was desired or not. If it is then correcting the proper 'next
> origin' is another trivial bug fix.
> 
> Some projects like single focused bug fixes for bisecting. I am fine
> with combining if that is wanted.

Agreed, two is better, I'll apply yours and add a followup fixing this
other issue, thanks.

- Arnaldo