perf inject: Remove stale build-id processing
I think we don't need to call build_id__mark_dso_hit() in the
perf_event__repipe_sample() as it's not used by -b option. In case of
the -b option is used, it uses perf_event__inject_buildid() instead.
This can remove unnecessary overhead of finding thread/map for each
sample event.
Also I suspect HEADER_BUILD_ID feature bit setting since we already
generated/injected BUILD_ID event into the output stream. So this
header information seems redundant. I'm not 100% sure about the
auxtrace usage, but it looks like not related to this directly.
And we now have --buildid-all so users can get the same behavior if
they want.
Cc: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
1 file changed
