[PATCH 1/6] perf tools: Synthetize the targeted process

[Frederic Weisbecker <fweisbec@gmail.com>]

Don't forget to also synthetize the targeted process from perf record
or we'll miss its dso in the events and then we won't be able to deal
with its build-id.

We are missing it because it is created after the existing synthetized
tasks but before the counters are enabled and can send its mapping
event.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
---
 tools/perf/builtin-record.c |   11 ++++++++++-
 1 files changed, 10 insertions(+), 1 deletions(-)

diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index cf2cd25..2b45d33 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -497,13 +497,22 @@ static int __cmd_record(int argc, const char **argv)
 	if (target_pid == -1 && argc) {
 		pid = fork();
 		if (pid < 0)
-			perror("failed to fork");
+			die("failed to fork");
 
 		if (!pid) {
 			if (execvp(argv[0], (char **)argv)) {
 				perror(argv[0]);
 				exit(-1);
 			}
+		} else {
+			/*
+			 * Wait a bit for the execv'ed child to appear
+			 * and be updated in /proc
+			 * FIXME: Do you know a less heuristical solution?
+			 */
+			usleep(1000);
+			event__synthesize_thread(pid,
+						 process_synthesized_event);
 		}
 
 		child_pid = pid;
-- 
1.6.2.3