[patch] current patch to catch when it fails (no manual delays)

printf_write_error_workaround_through_delay.patch (text/plain), 1.50 KB, created by abandoned account disabled email on 2015-10-17 04:36:43 UTC

left as is because it works, for now
that echo & ps delays seems to be preventing the write error from occurring so far

with this patch I just got a fail 1 for hwclock, so the delay must be not long enough to catch this all the time!

--- orig/sh/rc-cgroup.sh.in	2015-08-31 10:39:36.068009618 +0300
+++ patched/sh/rc-cgroup.sh.in	2015-08-31 10:38:29.817210595 +0300
@@ -85,8 +85,28 @@ cgroup_add_service()
 	openrc_cgroup=/sys/fs/cgroup/openrc
 	if [ -d "$openrc_cgroup" ]; then
 		cgroup="$openrc_cgroup/$RC_SVCNAME"
+
+    #the delay these introduce, seems to prevent wtw race is going on that causes this: /lib64/rc/sh/rc-cgroup.sh: line 89: printf: write error: No such device
+    #echo "Current: $$ $RC_SVCNAME" >/dev/null #2>/dev/null 1>&2
+    #ps afwj $$ >/dev/null #2>/dev/null 1>&2
+  #apparently fixed in openrc 0.18.3 old:^tested and confirmed to FAIL(2) (at boot) when the above 2 lines are commented out! but not when they are uncommented! so, draw your own conclusions.
 		mkdir -p "$cgroup"
-		[ -f "$cgroup/tasks" ] && printf "%d" 0 > "$cgroup/tasks"
+    local fail=0
+		if test -f "$cgroup/tasks"; then
+      if ! printf "%d" 0 > "$cgroup/tasks"; then
+        fail=1
+      fi
+    else
+      fail=2
+    fi
+    if test "$fail" -ne "0"; then
+      echo "------FAIL(${fail})----- $cgroup/tasks" 1>&2
+      echo "Current: $$ $RC_SVCNAME" 1>&2
+      cat "$cgroup/tasks"
+    fi
+#    [ -f "$cgroup/tasks" ] || ls -la "$openrc_cgroup" 1>&2
+
+#		[ -f "$cgroup/tasks" ] && printf "%d" 0 > "$cgroup/tasks"
 	fi
 }