Hi -
> [...]
> Frank, test 666 is leaking pmmgr and valgrind processes ...
> [...]
> I have two of those for every test run over the last few days. It
> looks like sudo's parent was killed, rather than the pmmgr child,
> since its been reparented to init.
ISTM the sudo is not necessary or helpful here. The following patch
takes that out, and doesn't affect results otherwise.
- FChE
diff --git a/qa/666 b/qa/666
index d6fbbadcd7cd..65bfed9d57ec 100755
--- a/qa/666
+++ b/qa/666
@@ -32,7 +32,7 @@ export PMCD_WAIT_TIMEOUT PMCD_CONNECT_TIMEOUT
PMCD_RECONNECT_TIMEOUT
_cleanup()
{
- if [ -n "$pid" ]; then $sudo kill $pid; fi
+ if [ -n "$pid" ]; then kill $pid; fi
# restart pmcd and primary pmlogger
$sudo $PCP_RC_DIR/pcp restart >>$seq.full 2>&1
_wait_for_pmcd
@@ -81,7 +81,7 @@ echo 'pcp://localhost:44321/' >>$tmp.dir/target-host
echo 'localhost6' >>$tmp.dir/target-host
echo 'pcp://localhost6:44321/' >>$tmp.dir/target-host
# note -v -v here is the same as -D appl0,appl1
-$sudo $_valgrind_clean_assert $PCP_BINADM_DIR/pmmgr -v -v -p 5 -l $tmp.out -c
$tmp.dir >$tmp.valout 2>$tmp.valerr &
+$_valgrind_clean_assert $PCP_BINADM_DIR/pmmgr -U `whoami` -v -v -p 5 -l
$tmp.out -c $tmp.dir >$tmp.valout 2>$tmp.valerr &
pid=$!
echo "pid=$!" >>$seq.full
@@ -227,7 +227,7 @@ done
date >>$seq.full
echo "=== ZZZ kill pmmgr ===" | tee -a $seq.full
-$sudo kill $pid
+kill $pid
pid=
sleep 2
|