sanitycheck: Fix process termination with newer ninja version
It appears that ninja 1.6.0 or greater don't seem to send SIGTERM down
to the child processes and thus we don't terminate correctly. This
causes a hang with renode simulations.
Change terminate call to 'self.try_kill_process_by_pid()' when test
state is updated (i.e. done running with either passed or failed), in
order to explicitly send a SIGTERM to the simulator process before
sending a SIGTERM to ninja.
Refactor the terminate code so we encapsulate the behavior in one place
for a BinaryHandler.
Based on change from Stephanos Ioannidis <root@stephanos.io>
Signed-off-by: Kumar Gala <kumar.gala@linaro.org>
diff --git a/scripts/sanitycheck b/scripts/sanitycheck
index ca4ebd0..cf2fcd2 100755
--- a/scripts/sanitycheck
+++ b/scripts/sanitycheck
@@ -537,6 +537,16 @@
except ProcessLookupError:
pass
+ def terminate(self, proc):
+ # encapsulate terminate functionality so we do it consistently where ever
+ # we might want to terminate the proc. We need try_kill_process_by_pid
+ # because of both how newer ninja (1.6.0 or greater) and .NET / renode
+ # work. Newer ninja's don't seem to pass SIGTERM down to the children
+ # so we need to use try_kill_process_by_pid.
+ self.try_kill_process_by_pid()
+ proc.terminate()
+ self.terminated = True
+
def _output_reader(self, proc, harness):
log_out_fp = open(self.log, "wt")
for line in iter(proc.stdout.readline, b''):
@@ -550,8 +560,7 @@
#so let's give it up to 100ms to do so
proc.wait(0.1)
except subprocess.TimeoutExpired:
- proc.terminate()
- self.terminated = True
+ self.terminate(proc)
break
log_out_fp.close()
@@ -596,9 +605,7 @@
t.start()
t.join(self.timeout)
if t.is_alive():
- self.try_kill_process_by_pid()
- proc.terminate()
- self.terminated = True
+ self.terminate(proc)
t.join()
proc.wait()
self.returncode = proc.returncode
