Re: [BUG] perf: buildid not managed properly when cmd path is relative

[Arnaldo Carvalho de Melo <acme@infradead.org>]

Em Tue, Jun 01, 2010 at 10:05:24PM +0200, Stephane Eranian escreveu:
> Hi,
> 
> You patch does seem to fix the problem.

thanks for confirming.
 
> However, I think there may be another issue, maybe
> caused by the patch.
> 
> If you monitor a program WITHOUT buildids, and you run
> 
> $ perf record tmp/foo
> [ perf record: Woken up 1 times to write data ]
> [ perf record: Captured and wrote 0.077 MB perf.data (~3370 samples) ]
> 
> $  ./perf buildid-list
> 86200ab2bca285344a55156a079debbbe172bcc5 [kernel.kallsyms]
> 
> foo does not appear, yet we know it got samples.
> 
> $ ./perf buildid-list --with-hits
> 86200ab2bca285344a55156a079debbbe172bcc5 [kernel.kallsyms]
> 0000000000000000000000000000000000000000 /home/eranian/tmp/foo
> 
> With the --with-hits option, I get more output than without.Supposedly
> no option means 'prints all entries', including those with hits or
> without buildids.

Right, I noticed that too, but concentrated on fixing the problem that
was clearly a bug, will add this to my TODO list as I agree this is
inconsistent and/or incomplete behaviour.

Will add a Tested-by: you and push it to Ingo via perf/urgent.

Thanks for testing,

- Arnaldo