Change simplenode template files to work on more platforms

When developing Riak, we have found bugs and other issues due
to the number of platforms we support.

Here is an overview of the changes:
  -  Fix command-line syntax for commands to work on *BSD / Sun
  -  Add chkconfig and getpid to nodetool
  -  Replace platform specific 'kill' commands with a nodetool
	getpid method
  -  Fix RUNNER_USER settings to work on *BSD

4 files changed, 156 insertions, 43 deletions

diff --git a/priv/templates/simplenode.erl.script b/priv/templates/simplenode.erl.script
index 6f65e3f..7919d69 100644
--- a/priv/templates/simplenode.erl.script
+++ b/priv/templates/simplenode.erl.script
@@ -1,13 +1,23 @@
 #!/bin/sh
 
-## This script replaces the default "erl" in erts-VSN/bin. This is necessary
-## as escript depends on erl and in turn, erl depends on having access to a
-## bootscript (start.boot). Note that this script is ONLY invoked as a side-effect
-## of running escript -- the embedded node bypasses erl and uses erlexec directly
-## (as it should).
+# /bin/sh on Solaris is not a POSIX compatible shell, but /usr/bin/ksh is.
+if [ `uname -s` = 'SunOS' -a "${POSIX_SHELL}" != "true" ]; then
+    POSIX_SHELL="true"
+    export POSIX_SHELL
+    exec /usr/bin/ksh $0 "$@"
+fi
+
+# clear it so if we invoke other scripts, they run as ksh as well
+unset POSIX_SHELL
+
+## This script replaces the default "erl" in erts-VSN/bin. This is
+## necessary as escript depends on erl and in turn, erl depends on
+## having access to a bootscript (start.boot). Note that this script
+## is ONLY invoked as a side-effect of running escript -- the embedded
+## node bypasses erl and uses erlexec directly (as it should).
 ##
-## Note that this script makes the assumption that there is a start_clean.boot
-## file available in $ROOTDIR/release/VSN.
+## Note that this script makes the assumption that there is a
+## start_clean.boot file available in $ROOTDIR/release/VSN.
 
 # Determine the abspath of where this script is executing from.
 ERTS_BIN_DIR=$(cd ${0%/*} && pwd)
diff --git a/priv/templates/simplenode.nodetool b/priv/templates/simplenode.nodetool
index eb08fa4..54ee6d6 100644..100755
--- a/priv/templates/simplenode.nodetool
+++ b/priv/templates/simplenode.nodetool
@@ -1,3 +1,4 @@
+#!/usr/bin/env escript
 %% -*- mode: erlang;erlang-indent-level: 4;indent-tabs-mode: nil -*-
 %% ex: ft=erlang ts=4 sw=4 et
 %% -------------------------------------------------------------------
@@ -5,25 +6,53 @@
 %% nodetool: Helper Script for interacting with live nodes
 %%
 %% -------------------------------------------------------------------
+-mode(compile).
 
 main(Args) ->
     ok = start_epmd(),
     %% Extract the args
     {RestArgs, TargetNode} = process_args(Args, [], undefined),
 
+    %% any commands that don't need a running node
+    case RestArgs of
+        ["chkconfig", File] ->
+            case file:consult(File) of
+                {ok, _} ->
+                    io:format("ok\n"),
+                    halt(0);
+                {error, {Line, Mod, Term}} ->
+                    io:format(standard_error, ["Error on line ",
+                              file:format_error({Line, Mod, Term}), "\n"], []),
+                    halt(1);
+                {error, R} ->
+                    io:format(standard_error, ["Error reading config file: ",
+                              file:format_error(R), "\n"], []),
+                    halt(1)
+            end;
+        _ ->
+            ok
+    end,
+
     %% See if the node is currently running  -- if it's not, we'll bail
-    case {net_kernel:hidden_connect_node(TargetNode), net_adm:ping(TargetNode)} of
+    case {net_kernel:hidden_connect_node(TargetNode),
+          net_adm:ping(TargetNode)} of
         {true, pong} ->
             ok;
+        {false,pong} ->
+            io:format("Failed to connect to node ~p .\n", [TargetNode]),
+            halt(1);
         {_, pang} ->
             io:format("Node ~p not responding to pings.\n", [TargetNode]),
             halt(1)
     end,
 
     case RestArgs of
+        ["getpid"] ->
+            io:format("~p\n",
+                      [list_to_integer(rpc:call(TargetNode, os, getpid, []))]);
         ["ping"] ->
-            %% If we got this far, the node already responsed to a ping, so just dump
-            %% a "pong"
+            %% If we got this far, the node already responsed to a
+            %% ping, so just dump a "pong"
             io:format("pong\n");
         ["stop"] ->
             io:format("~p\n", [rpc:call(TargetNode, init, stop, [], 60000)]);
@@ -32,7 +61,9 @@ main(Args) ->
         ["reboot"] ->
             io:format("~p\n", [rpc:call(TargetNode, init, reboot, [], 60000)]);
         ["rpc", Module, Function | RpcArgs] ->
-            case rpc:call(TargetNode, list_to_atom(Module), list_to_atom(Function),
+            case rpc:call(TargetNode,
+                          list_to_atom(Module),
+                          list_to_atom(Function),
                           [RpcArgs], 60000) of
                 ok ->
                     ok;
@@ -42,8 +73,23 @@ main(Args) ->
                 _ ->
                     halt(1)
             end;
+        ["rpc_infinity", Module, Function | RpcArgs] ->
+            case rpc:call(TargetNode,
+                          list_to_atom(Module),
+                          list_to_atom(Function),
+                          [RpcArgs], infinity) of
+                ok ->
+                    ok;
+                {badrpc, Reason} ->
+                    io:format("RPC to ~p failed: ~p\n", [TargetNode, Reason]),
+                    halt(1);
+                _ ->
+                    halt(1)
+            end;
         ["rpcterms", Module, Function, ArgsAsString] ->
-            case rpc:call(TargetNode, list_to_atom(Module), list_to_atom(Function),
+            case rpc:call(TargetNode,
+                          list_to_atom(Module),
+                          list_to_atom(Function),
                           consult(ArgsAsString), 60000) of
                 {badrpc, Reason} ->
                     io:format("RPC to ~p failed: ~p\n", [TargetNode, Reason]),
@@ -53,7 +99,7 @@ main(Args) ->
             end;
         Other ->
             io:format("Other: ~p\n", [Other]),
-            io:format("Usage: nodetool {ping|stop|restart|reboot}\n")
+            io:format("Usage: nodetool {chkconfig|getpid|ping|stop|restart|reboot|rpc|rpc_infinity|rpcterms}\n")
     end,
     net_kernel:stop().
 
diff --git a/priv/templates/simplenode.reltool.config b/priv/templates/simplenode.reltool.config
index b580c2a..4189329 100644
--- a/priv/templates/simplenode.reltool.config
+++ b/priv/templates/simplenode.reltool.config
@@ -1,3 +1,5 @@
+%% -*- mode: erlang -*-
+%% ex: ft=erlang
 {sys, [
        {lib_dirs, []},
        {erts, [{mod_cond, derived}, {app_file, strip}]},
diff --git a/priv/templates/simplenode.runner b/priv/templates/simplenode.runner
index 43d90bc..8c55e3d 100755
--- a/priv/templates/simplenode.runner
+++ b/priv/templates/simplenode.runner
@@ -2,6 +2,19 @@
 # -*- tab-width:4;indent-tabs-mode:nil -*-
 # ex: ts=4 sw=4 et
 
+# /bin/sh on Solaris is not a POSIX compatible shell, but /usr/bin/ksh is.
+if [ `uname -s` = 'SunOS' -a "${POSIX_SHELL}" != "true" ]; then
+    POSIX_SHELL="true"
+    export POSIX_SHELL
+    # To support 'whoami' add /usr/ucb to path
+    PATH=/usr/ucb:$PATH
+    export PATH
+    exec /usr/bin/ksh $0 "$@"
+fi
+
+# clear it so if we invoke other scripts, they run as ksh
+unset POSIX_SHELL
+
 RUNNER_SCRIPT_DIR=$(cd ${0%/*} && pwd)
 
 CALLER_DIR=$PWD
@@ -11,10 +24,17 @@ RUNNER_ETC_DIR=$RUNNER_BASE_DIR/etc
 # Note the trailing slash on $PIPE_DIR/
 PIPE_DIR=/tmp/$RUNNER_BASE_DIR/
 RUNNER_USER=
+WHOAMI=$(whoami)
 
 # Make sure this script is running as the appropriate user
-if [ ! -z "$RUNNER_USER" ] && [ `whoami` != "$RUNNER_USER" ]; then
-    exec sudo -u $RUNNER_USER -i $0 $@
+if ([ "$RUNNER_USER" ] && [ "x$WHOAMI" != "x$RUNNER_USER" ]); then
+    type sudo > /dev/null 2>&1
+    if [ $? -ne 0 ]; then
+        echo "sudo doesn't appear to be installed and your EUID isn't $RUNNER_USER" 1>&2
+        exit 1
+    fi
+    echo "Attempting to restart script through sudo -H -u $RUNNER_USER" >&2
+    exec sudo -H -u $RUNNER_USER -i $RUNNER_SCRIPT_DIR/$RUNNER_SCRIPT $@
 fi
 
 # Identify the script name
@@ -25,7 +45,8 @@ START_ERL=`cat $RUNNER_BASE_DIR/releases/start_erl.data`
 ERTS_VSN=${START_ERL% *}
 APP_VSN=${START_ERL#* }
 
-# Use $CWD/vm.args if exists, otherwise releases/APP_VSN/vm.args, or else etc/vm.args
+# Use $CWD/vm.args if exists, otherwise releases/APP_VSN/vm.args, or
+# else etc/vm.args
 if [ -e "$CALLER_DIR/vm.args" ]; then
     VMARGS_PATH=$CALLER_DIR/vm.args
     USE_DIR=$CALLER_DIR
@@ -54,7 +75,7 @@ else
 fi
 
 # Extract the target node name from node.args
-NAME_ARG=`egrep '^-s?name' $VMARGS_PATH`
+NAME_ARG=`egrep '^\-s?name' $VMARGS_PATH`
 if [ -z "$NAME_ARG" ]; then
     echo "vm.args needs to have either -name or -sname parameter."
     exit 1
@@ -64,12 +85,13 @@ fi
 REMSH_TYPE=`echo $NAME_ARG | awk '{print $1}'`
 REMSH_NAME=`echo $NAME_ARG | awk '{print $2}'`
 
-# Note the `date +%s`, used to allow multiple remsh to the same node transparently
+# Note the `date +%s`, used to allow multiple remsh to the same node
+# transparently
 REMSH_NAME_ARG="$REMSH_TYPE remsh`date +%s`@`echo $REMSH_NAME | awk -F@ '{print $2}'`"
 REMSH_REMSH_ARG="-remsh $REMSH_NAME"
 
 # Extract the target cookie
-COOKIE_ARG=`grep '^-setcookie' $VMARGS_PATH`
+COOKIE_ARG=`grep '^\-setcookie' $VMARGS_PATH`
 if [ -z "$COOKIE_ARG" ]; then
     echo "vm.args needs to have a -setcookie parameter."
     exit 1
@@ -81,7 +103,6 @@ cd $USE_DIR
 # Make sure log directory exists
 mkdir -p $USE_DIR/log
 
-
 # Add ERTS bin dir to our path
 ERTS_PATH=$RUNNER_BASE_DIR/erts-$ERTS_VSN/bin
 
@@ -91,11 +112,35 @@ NODETOOL="$ERTS_PATH/escript $ERTS_PATH/nodetool $NAME_ARG $COOKIE_ARG"
 # Setup remote shell command to control node
 REMSH="$ERTS_PATH/erl $REMSH_NAME_ARG $REMSH_REMSH_ARG $COOKIE_ARG"
 
+# Common functions
+
+# Ping node without allowing nodetool to take stdin
+ping_node() {
+    $NODETOOL ping < /dev/null
+}
+
+# Set the PID global variable, return 1 on error
+get_pid() {
+    PID=`$NODETOOL getpid < /dev/null`
+    ES=$?
+    if [ "$ES" -ne 0 ]; then
+        echo "Node is not running!"
+        return 1
+    fi
+
+    # don't allow empty or init pid's
+    if [ -z $PID ] || [ "$PID" -le 1 ]; then
+        return 1
+    fi
+
+    return 0
+}
+
 # Check the first argument for instructions
 case "$1" in
     start|start_boot)
         # Make sure there is not already a node running
-        RES=`$NODETOOL ping`
+        RES=`ping_node`
         if [ "$RES" = "pong" ]; then
             echo "Node is already running!"
             exit 1
@@ -122,27 +167,28 @@ case "$1" in
     stop)
         # Wait for the node to completely stop...
         case `uname -s` in
-            Linux|Darwin|FreeBSD|DragonFly|NetBSD|OpenBSD)
-                # PID COMMAND
-                PID=`ps ax -o pid= -o command=|\
-                    grep "$RUNNER_BASE_DIR/.*/[b]eam"|awk '{print $1}'`
-                ;;
-            SunOS)
-                # PID COMMAND
-                PID=`ps -ef -o pid= -o args=|\
-                    grep "$RUNNER_BASE_DIR/.*/[b]eam"|awk '{print $1}'`
-                ;;
-            CYGWIN*)
-                # UID PID PPID TTY STIME COMMAND
-                PID=`ps -efW|grep "$RUNNER_BASE_DIR/.*/[b]eam"|awk '{print $2}'`
-                ;;
+            Darwin)
+                # Make sure we explicitly set this because iTerm.app doesn't for
+                # some reason.
+                COMMAND_MODE=unix2003
         esac
+
+        # Get the PID from nodetool
+        get_pid
+        GPR=$?
+        if [ "$GPR" -ne 0 ] || [ -z $PID ]; then
+            exit $GPR
+        fi
+
+        # Tell nodetool to initiate a stop
         $NODETOOL stop
         ES=$?
         if [ "$ES" -ne 0 ]; then
             exit $ES
         fi
-        while `kill -0 $PID 2>/dev/null`;
+
+        # Wait for the node to completely stop...
+        while `kill -s 0 $PID 2>/dev/null`
         do
             sleep 1
         done
@@ -168,7 +214,7 @@ case "$1" in
 
     ping)
         ## See if the VM is alive
-        $NODETOOL ping
+        ping_node
         ES=$?
         if [ "$ES" -ne 0 ]; then
             exit $ES
@@ -176,8 +222,8 @@ case "$1" in
         ;;
 
     attach)
-        # Make sure a node IS running
-        RES=`$NODETOOL ping`
+        # Make sure a node is running
+        ping_node
         ES=$?
         if [ "$ES" -ne 0 ]; then
             echo "Node is not running!"
@@ -189,8 +235,8 @@ case "$1" in
         ;;
 
     remote_console)
-        # Make sure a node IS running
-        RES=`$NODETOOL ping`
+        # Make sure a node is running
+        ping_node
         ES=$?
         if [ "$ES" -ne 0 ]; then
             echo "Node is not running!"
@@ -210,7 +256,7 @@ case "$1" in
         fi
 
         # Make sure a node IS running
-        RES=`$NODETOOL ping`
+        ping_node
         ES=$?
         if [ "$ES" -ne 0 ]; then
             echo "Node is not running!"
@@ -283,8 +329,17 @@ case "$1" in
         # Start the VM
         exec $CMD -- ${1+"$@"}
         ;;
+    getpid)
+        # Get the PID from nodetool
+        get_pid
+        ES=$?
+        if [ "$ES" -ne 0 ] || [ -z $PID ]; then
+            exit $ES
+        fi
+        echo $PID
+        ;;
     *)
-        echo "Usage: $SCRIPT {start|start_boot <file>|foreground|stop|restart|reboot|ping|console|console_clean|console_boot <file>|attach|remote_console|upgrade}"
+        echo "Usage: $SCRIPT {start|start_boot <file>|foreground|stop|restart|reboot|ping|console|getpid||console_clean|console_boot <file>|attach|remote_console|upgrade}"
         exit 1
         ;;
 esac