Global destructor clean-ups and python test fixes

diff --git a/handle.cc b/handle.cc
index 926cb59..20438f9 100644 (file)
--- a/handle.cc
+++ b/handle.cc
@@ -12,6 +12,16 @@ public:
 static std::mutex mgr_mutex;
 static std::map<string, shared_ptr<hinfo>> hmap;
 
 static std::mutex mgr_mutex;
 static std::map<string, shared_ptr<hinfo>> hmap;
 

+void handle::shutdown() {
+    lock ml(mgr_mutex);
+    LOG_NONMEMBER << "Shutting down handle manager";
+    for (auto p : hmap) {
+        p.second->valid = false;
+        LOG_NONMEMBER << "cl " << p.first << " refcnt " << p.second.use_count();
+    }
+    hmap.clear();
+}
+

 handle::handle(const string & destination) : destination_(destination) {
     lock ml(mgr_mutex);
     h = hmap[destination];
 handle::handle(const string & destination) : destination_(destination) {
     lock ml(mgr_mutex);
     h = hmap[destination];

diff --git a/handle.h b/handle.h
index 92110bd..e569002 100644 (file)
--- a/handle.h
+++ b/handle.h
@@ -26,6 +26,7 @@ class handle {
         handle(const string & destination);
         rpcc *safebind();
         void invalidate();
         handle(const string & destination);
         rpcc *safebind();
         void invalidate();

+        static void shutdown();

 };
 
 #endif
 };
 
 #endif

diff --git a/lock_client.cc b/lock_client.cc
index 9949eac..009e051 100644 (file)
--- a/lock_client.cc
+++ b/lock_client.cc
@@ -44,10 +44,22 @@ lock_client::lock_client(string xdst, lock_release_user *_lu) : lu(_lu), next_xi
     rlsrpc->start();
 }
 
     rlsrpc->start();
 }
 

+lock_client::~lock_client() {
+    release_fifo.enq(nothing<lock_protocol::lockid_t>());
+    releaser_thread.join();
+}
+

 void lock_client::releaser() {
     while (1) {
 void lock_client::releaser() {
     while (1) {

-        lock_protocol::lockid_t lid;
-        release_fifo.deq(&lid);
+        maybe<lock_protocol::lockid_t> mlid;
+        release_fifo.deq(&mlid);
+
+        if (!mlid) {
+            LOG << "Releaser stopping";
+            break;
+        }
+
+        lock_protocol::lockid_t lid = mlid;

         LOG << "Releaser: " << lid;
 
         lock_state & st = get_lock_state(lid);
         LOG << "Releaser: " << lid;
 
         lock_state & st = get_lock_state(lid);

@@ -122,7 +134,7 @@ lock_protocol::status lock_client::acquire(lock_protocol::lockid_t lid) {
                 st.state = lock_state::locked;
                 st.held_by = releaser_thread.get_id();
                 LOG << "Queuing " << lid << " for release";
                 st.state = lock_state::locked;
                 st.held_by = releaser_thread.get_id();
                 LOG << "Queuing " << lid << " for release";

-                release_fifo.enq(lid);
+                release_fifo.enq(just(lid));

             } else if (front == self) {
                 st.wanted_by.pop_front();
                 st.state = lock_state::locked;
             } else if (front == self) {
                 st.wanted_by.pop_front();
                 st.state = lock_state::locked;

@@ -156,7 +168,7 @@ lock_protocol::status lock_client::release(lock_protocol::lockid_t lid) {
             st.held_by = releaser_thread.get_id();
             st.wanted_by.pop_front();
             LOG << "Queuing " << lid << " for release";
             st.held_by = releaser_thread.get_id();
             st.wanted_by.pop_front();
             LOG << "Queuing " << lid << " for release";

-            release_fifo.enq(lid);
+            release_fifo.enq(just(lid));

         } else
             st.signal(front);
     }
         } else
             st.signal(front);
     }

@@ -179,7 +191,7 @@ rlock_protocol::status lock_client::revoke_handler(int &, lock_protocol::lockid_
         st.held_by = releaser_thread.get_id();
         if (st.wanted_by.size())
             st.wanted_by.pop_front();
         st.held_by = releaser_thread.get_id();
         if (st.wanted_by.size())
             st.wanted_by.pop_front();

-        release_fifo.enq(lid);
+        release_fifo.enq(just(lid));

     } else {
         // get in line
         st.wanted_by.push_back(releaser_thread.get_id());
     } else {
         // get in line
         st.wanted_by.push_back(releaser_thread.get_id());

diff --git a/lock_client.h b/lock_client.h
index 5cdb26e..74f17a8 100644 (file)
--- a/lock_client.h
+++ b/lock_client.h
@@ -9,6 +9,7 @@
 #include "lock_protocol.h"
 #include "rpc/fifo.h"
 #include "rsm_client.h"
 #include "lock_protocol.h"
 #include "rpc/fifo.h"
 #include "rsm_client.h"

+#include "maybe.h"

 
 class lock_release_user {
     public:
 
 class lock_release_user {
     public:

@@ -52,18 +53,18 @@ class lock_client {
         string id;
         std::mutex xid_mutex;
         lock_protocol::xid_t next_xid;
         string id;
         std::mutex xid_mutex;
         lock_protocol::xid_t next_xid;

-        fifo<lock_protocol::lockid_t> release_fifo;
+        fifo<maybe<lock_protocol::lockid_t>> release_fifo;

         std::mutex lock_table_lock;
         lock_map lock_table;
         lock_state & get_lock_state(lock_protocol::lockid_t lid);
     public:
         static in_port_t last_port;
         lock_client(string xdst, lock_release_user *l = 0);
         std::mutex lock_table_lock;
         lock_map lock_table;
         lock_state & get_lock_state(lock_protocol::lockid_t lid);
     public:
         static in_port_t last_port;
         lock_client(string xdst, lock_release_user *l = 0);

-        ~lock_client() {}
+        ~lock_client();

         lock_protocol::status acquire(lock_protocol::lockid_t);
         lock_protocol::status release(lock_protocol::lockid_t);
         int stat(lock_protocol::lockid_t);
         lock_protocol::status acquire(lock_protocol::lockid_t);
         lock_protocol::status release(lock_protocol::lockid_t);
         int stat(lock_protocol::lockid_t);

-        void releaser NORETURN ();
+        void releaser();

         rlock_protocol::status revoke_handler(int &, lock_protocol::lockid_t, lock_protocol::xid_t);
         rlock_protocol::status retry_handler(int &, lock_protocol::lockid_t, lock_protocol::xid_t);
 };
         rlock_protocol::status revoke_handler(int &, lock_protocol::lockid_t, lock_protocol::xid_t);
         rlock_protocol::status retry_handler(int &, lock_protocol::lockid_t, lock_protocol::xid_t);
 };

diff --git a/lock_tester.cc b/lock_tester.cc
index 7143401..58e691f 100644 (file)
--- a/lock_tester.cc
+++ b/lock_tester.cc
@@ -5,13 +5,14 @@
 #include "lock_client.h"
 #include <arpa/inet.h>
 #include <unistd.h>
 #include "lock_client.h"
 #include <arpa/inet.h>
 #include <unistd.h>

+#include "handle.h"

 
 char log_thread_prefix = 'c';
 
 // must be >= 2
 const int nt = 6; //XXX: lab1's rpc handlers are blocking. Since rpcs uses a thread pool of 10 threads, we cannot test more than 10 blocking rpc.
 static string dst;
 
 char log_thread_prefix = 'c';
 
 // must be >= 2
 const int nt = 6; //XXX: lab1's rpc handlers are blocking. Since rpcs uses a thread pool of 10 threads, we cannot test more than 10 blocking rpc.
 static string dst;

-static lock_client **lc = new lock_client * [nt];
+static lock_client *lc[nt];

 static lock_protocol::lockid_t a = "1";
 static lock_protocol::lockid_t b = "2";
 static lock_protocol::lockid_t c = "3";
 static lock_protocol::lockid_t a = "1";
 static lock_protocol::lockid_t b = "2";
 static lock_protocol::lockid_t c = "3";

@@ -180,4 +181,11 @@ main(int argc, char *argv[])
 
     LOG_NONMEMBER << argv[0] << ": passed all tests successfully";
 
 
     LOG_NONMEMBER << argv[0] << ": passed all tests successfully";
 

+    for (int i = 0; i < nt; i++)
+        delete lc[i];
+
+    handle::shutdown();
+    poll_mgr::shared_mgr.shutdown();
+
+    LOG_NONMEMBER << argv[0] << ": clean-up complete";

 }
 }

diff --git a/maybe.h b/maybe.h
new file mode 100644 (file)
index 0000000..565949e
--- /dev/null
+++ b/maybe.h
@@ -0,0 +1,32 @@
+#ifndef maybe_h
+#define maybe_h
+
+#include <cassert>
+
+template <class T> struct maybe {
+    bool isJust = false;
+    union { T t; };
+    maybe() {}
+    maybe(const maybe & other) {
+        *this = other;
+    }
+    ~maybe() {
+        if (isJust)
+            t.~T();
+    }
+    const maybe & operator =(const maybe & other) {
+        if (isJust) t.~T();
+        isJust = other.isJust;
+        if(isJust) new(&t) T(other.t);
+        return *this;
+    }
+    operator T &() { assert(isJust); return t; }
+    operator bool() { return isJust; }
+};
+
+template <class T>
+static inline maybe<T> nothing() {return maybe<T>();}
+template <class T>
+static inline maybe<T> just(const T & t_) { maybe<T> m; m.isJust = true; new(&m.t) T(t_); return m; }
+
+#endif

diff --git a/rpc/poll_mgr.cc b/rpc/poll_mgr.cc
index 83289a8..ebd61e1 100644 (file)
--- a/rpc/poll_mgr.cc
+++ b/rpc/poll_mgr.cc
@@ -59,9 +59,14 @@ poll_mgr::poll_mgr() : aio_(new SelectAIO()) {
     th_ = thread(&poll_mgr::wait_loop, this);
 }
 
     th_ = thread(&poll_mgr::wait_loop, this);
 }
 

-poll_mgr::~poll_mgr()
-{
+poll_mgr::~poll_mgr() {
+    shutdown();
+}
+
+void poll_mgr::shutdown() {

     lock ml(m_);
     lock ml(m_);

+    if (shutdown_)
+        return;

     for (auto p : callbacks_)
         aio_->unwatch_fd(p.first, CB_RDWR);
     pending_change_ = true;
     for (auto p : callbacks_)
         aio_->unwatch_fd(p.first, CB_RDWR);
     pending_change_ = true;

@@ -71,9 +76,7 @@ poll_mgr::~poll_mgr()
     th_.join();
 }
 
     th_.join();
 }
 

-void
-poll_mgr::add_callback(int fd, poll_flag flag, aio_callback *ch)
-{
+void poll_mgr::add_callback(int fd, poll_flag flag, aio_callback *ch) {

     lock ml(m_);
     aio_->watch_fd(fd, flag);
 
     lock ml(m_);
     aio_->watch_fd(fd, flag);
 

diff --git a/rpc/poll_mgr.h b/rpc/poll_mgr.h
index d8cfd20..babf908 100644 (file)
--- a/rpc/poll_mgr.h
+++ b/rpc/poll_mgr.h
@@ -31,6 +31,7 @@ class poll_mgr {
         void del_callback(int fd, poll_flag flag);
         void block_remove_fd(int fd);
         void wait_loop();
         void del_callback(int fd, poll_flag flag);
         void block_remove_fd(int fd);
         void wait_loop();

+        void shutdown();

 
     private:
         std::mutex m_;
 
     private:
         std::mutex m_;

diff --git a/rsm_tester.py b/rsm_tester.py
index 32f0ad2..c7b8bbd 100755 (executable)
--- a/rsm_tester.py
+++ b/rsm_tester.py
@@ -29,7 +29,7 @@ for sig in ['HUP', 'INT', 'ABRT', 'QUIT', 'TERM']:
     signal.signal(num, killprocess)
 
 def paxos_log(port):
     signal.signal(num, killprocess)
 
 def paxos_log(port):

-  return "paxos-%d.log" % port
+    return "paxos-%d.log" % port

 
 def die(*s):
     print >>sys.stderr, ''.join(s)
 
 def die(*s):
     print >>sys.stderr, ''.join(s)

@@ -77,7 +77,7 @@ def spawn(p, *a):
         try:
             os.execv(p, [p] + sa)
         except OSError, e:
         try:
             os.execv(p, [p] + sa)
         except OSError, e:

-            mydie("Cannot start new %s %s %s", (p, repr(sa), repr(e)))
+            mydie("Cannot start new %s %s %s" % (p, repr(sa), repr(e)))

 
 def randports(num):
     return sorted([random.randint(0, 54000/2)*2+10000 for i in xrange(num)])
 
 def randports(num):
     return sorted([random.randint(0, 54000/2)*2+10000 for i in xrange(num)])

@@ -114,13 +114,13 @@ def check_views(l, vs, last_v=None):
             # let there be extra views
             continue
         expected = vs[i]
             # let there be extra views
             continue
         expected = vs[i]

-        if tuple(expected) != tuple(view):
+        if set(expected) != set(view):

             mydie("Failed: In log %s at view %s is (%s), but expected %s (%s)" %
                   (l, str(num), repr(view), str(i), repr(expected)))
         i+=1
     if i < len(vs):
         mydie("Failed: In log %s, not enough views seen!" % (l,))
             mydie("Failed: In log %s at view %s is (%s), but expected %s (%s)" %
                   (l, str(num), repr(view), str(i), repr(expected)))
         i+=1
     if i < len(vs):
         mydie("Failed: In log %s, not enough views seen!" % (l,))

-    if last_v is not None and tuple(last_v) != tuple(last_view):
+    if last_v is not None and set(last_v) != set(last_view):

         mydie("Failed: In log %s last view didn't match, got view %s, but expected %s" %
               (l, repr(last_view), repr(last_v)))
 
         mydie("Failed: In log %s last view didn't match, got view %s, but expected %s" %
               (l, repr(last_view), repr(last_v)))
 

@@ -155,10 +155,10 @@ def wait_for_view_change(log, num_views, including, timeout):
 
 def waitpid_to(pid, to):
     start = time.time()
 
 def waitpid_to(pid, to):
     start = time.time()

-    done_pid = -1
-    while done_pid <= 0 and (time.time() - start) < to:
-      usleep(100000)
-      done_pid = os.waitpid(pid, os.WNOHANG)
+    done_pid = (0,0)
+    while done_pid == (0,0) and (time.time() - start) < to:
+        usleep(100000)
+        done_pid = os.waitpid(pid, os.WNOHANG)

 
     if done_pid <= 0:
         os.kill(pid, signal.SIGKILL)
 
     if done_pid <= 0:
         os.kill(pid, signal.SIGKILL)

@@ -240,7 +240,7 @@ if do_run[1]:
 if do_run[2]:
     print "test2: start 3-process lock server, kill first server"
     start_nodes(3,"ls")
 if do_run[2]:
     print "test2: start 3-process lock server, kill first server"
     start_nodes(3,"ls")

-    print "Kill first (PID: $pid[0]) on port $p[0]"
+    print "Kill first (PID: %d) on port %d" % (pid[0], p[0])

     os.kill(pid[0], signal.SIGTERM)
     usleep(500000)
     # it should go through 4 views
     os.kill(pid[0], signal.SIGTERM)
     usleep(500000)
     # it should go through 4 views

@@ -252,12 +252,12 @@ if do_run[2]:
 if do_run[3]:
     print "test3: start 3-process lock_server, kill a server, restart a server"
     start_nodes(3,"ls")
 if do_run[3]:
     print "test3: start 3-process lock_server, kill a server, restart a server"
     start_nodes(3,"ls")

-    print "Kill server (PID: $pid[2]) on port $p[2]"
+    print "Kill server (PID: %s) on port %s" % (pid[2], p[2])

     os.kill(pid[2], signal.SIGTERM)
     usleep(500000)
     v4 = (p[0], p[1])
     wait_and_check_expected_view(v4)
     os.kill(pid[2], signal.SIGTERM)
     usleep(500000)
     v4 = (p[0], p[1])
     wait_and_check_expected_view(v4)

-    print "Restart killed server on port $p[2]"
+    print "Restart killed server on port %s" % (p[2],)

     pid[2] = spawn_ls (p[0], p[2])
     usleep(500000)
     v5 = (p[0], p[1], p[2])
     pid[2] = spawn_ls (p[0], p[2])
     usleep(500000)
     v5 = (p[0], p[1], p[2])

@@ -268,34 +268,34 @@ if do_run[3]:
 if do_run[4]:
     print "test4: 3-process lock_server, kill third server, kill second server, restart third server, kill third server again, restart second server, re-restart third server, check logs"
     start_nodes(3,"ls")
 if do_run[4]:
     print "test4: 3-process lock_server, kill third server, kill second server, restart third server, kill third server again, restart second server, re-restart third server, check logs"
     start_nodes(3,"ls")

-    print "Kill server (PID: $pid[2]) on port $p[2]"
+    print "Kill server (PID: %s) on port %s" % (pid[2], p[2])

     os.kill(pid[2], signal.SIGTERM)
     usleep(500000)
     v4 = (p[0], p[1])
     wait_and_check_expected_view(v4)
     os.kill(pid[2], signal.SIGTERM)
     usleep(500000)
     v4 = (p[0], p[1])
     wait_and_check_expected_view(v4)

-    print "Kill server (PID: $pid[1]) on port $p[1]"
+    print "Kill server (PID: %s) on port %s" % (pid[1], p[1])

     os.kill(pid[1], signal.SIGTERM)
     usleep(500000)
     #no view change can happen because of a lack of majority
     os.kill(pid[1], signal.SIGTERM)
     usleep(500000)
     #no view change can happen because of a lack of majority

-    print "Restarting server on port $p[2]"
+    print "Restarting server on port %s" % (p[2],)

     pid[2] = spawn_ls(p[0], p[2])
     usleep(500000)
     #no view change can happen because of a lack of majority
     for port in p[0:1+2]:
         num_v = get_num_views(paxos_log(port), port)
         if num_v != in_views[port]:
     pid[2] = spawn_ls(p[0], p[2])
     usleep(500000)
     #no view change can happen because of a lack of majority
     for port in p[0:1+2]:
         num_v = get_num_views(paxos_log(port), port)
         if num_v != in_views[port]:

-            die("$num_v views in ", paxos_log(port), " : no new views should be formed due to the lack of majority")
+            die("%s_v views in ", paxos_log(port), " : no new views should be formed due to the lack of majority" % (num,))

     # kill node 3 again,
     # kill node 3 again,

-    print "Kill server (PID: $pid[2]) on port $p[2]"
+    print "Kill server (PID: %s) on port %s" % (pid[2], p[2])

     os.kill(pid[2], signal.SIGTERM)
     usleep(500000)
     os.kill(pid[2], signal.SIGTERM)
     usleep(500000)

-    print "Restarting server on port $p[1]"
+    print "Restarting server on port %s" % (p[1],)

     pid[1] = spawn_ls(p[0], p[1])
     usleep(700000)
     for port in p[0:1+1]:
         in_views[port] = get_num_views(paxos_log(port), port)
     pid[1] = spawn_ls(p[0], p[1])
     usleep(700000)
     for port in p[0:1+1]:
         in_views[port] = get_num_views(paxos_log(port), port)

-        print "   Node $port is present in ", in_views[port], " views in ", paxos_log(port), ""
-    print "Restarting server on port $p[2]"
+        print "   Node %s is present in " % (port,), in_views[port], " views in ", paxos_log(port)
+    print "Restarting server on port %s" % (p[2],)

     pid[2] = spawn_ls(p[0], p[2])
     lastv = (p[0],p[1],p[2])
     for port in lastv:
     pid[2] = spawn_ls(p[0], p[2])
     lastv = (p[0],p[1],p[2])
     for port in lastv:

@@ -309,17 +309,17 @@ if do_run[4]:
 if do_run[5]:
     print "test5: 3-process lock_server, send signal 1 to first server, kill third server, restart third server, check logs"
     start_nodes(3,"ls")
 if do_run[5]:
     print "test5: 3-process lock_server, send signal 1 to first server, kill third server, restart third server, check logs"
     start_nodes(3,"ls")

-    print "Sending paxos breakpoint 1 to first server on port $p[0]"
+    print "Sending paxos breakpoint 1 to first server on port %s" % (p[0],)

     spawn("./rsm_tester", p[0]+1, "breakpoint", 3)
     usleep(100000)
     spawn("./rsm_tester", p[0]+1, "breakpoint", 3)
     usleep(100000)

-    print "Kill third server (PID: $pid[2]) on port $p[2]"
+    print "Kill third server (PID: %s) on port %s" % (pid[2], p[2])

     os.kill(pid[2], signal.SIGTERM)
     usleep(500000)
     for port in p[0:1+2]:
         num_v = get_num_views(paxos_log(port), port)
         if num_v != in_views[port]:
     os.kill(pid[2], signal.SIGTERM)
     usleep(500000)
     for port in p[0:1+2]:
         num_v = get_num_views(paxos_log(port), port)
         if num_v != in_views[port]:

-            die("$num_v views in ", paxos_log(port), " : no new views should be formed due to the lack of majority")
-    print "Restarting third server on port $p[2]"
+            die("%s_v views in ", paxos_log(port), " : no new views should be formed due to the lack of majority" % (num,))
+    print "Restarting third server on port %s" % (p[2],)

     pid[2]= spawn_ls(p[0], p[2])
     lastv = (p[1],p[2])
     for port in lastv:
     pid[2]= spawn_ls(p[0], p[2])
     lastv = (p[1],p[2])
     for port in lastv:

@@ -334,18 +334,18 @@ if do_run[5]:
 if do_run[6]:
     print "test6: 4-process lock_server, send signal 2 to first server, kill fourth server, restart fourth server, check logs"
     start_nodes(4,"ls")
 if do_run[6]:
     print "test6: 4-process lock_server, send signal 2 to first server, kill fourth server, restart fourth server, check logs"
     start_nodes(4,"ls")

-    print "Sending paxos breakpoint 2 to first server on port $p[0]"
+    print "Sending paxos breakpoint 2 to first server on port %s" % (p[0],)

     spawn("./rsm_tester", p[0]+1, "breakpoint", 4)
     usleep(100000)
     spawn("./rsm_tester", p[0]+1, "breakpoint", 4)
     usleep(100000)

-    print "Kill fourth server (PID: $pid[3]) on port $p[3]"
+    print "Kill fourth server (PID: %s) on port %s" % (pid[3], p[3])

     os.kill(pid[3], signal.SIGTERM)
     usleep(500000)
     for port in (p[1],p[2]):
         num_v = get_num_views(paxos_log(port), port)
         if num_v != in_views[port]:
     os.kill(pid[3], signal.SIGTERM)
     usleep(500000)
     for port in (p[1],p[2]):
         num_v = get_num_views(paxos_log(port), port)
         if num_v != in_views[port]:

-            die("$num_v views in ", paxos_log(port), " : no new views should be formed due to the lack of majority")
+            die("%s_v views in ", paxos_log(port), " : no new views should be formed due to the lack of majority" % (num,))

     usleep(500000)
     usleep(500000)

-    print "Restarting fourth server on port $p[3]"
+    print "Restarting fourth server on port %s" % (p[3],)

     pid[3] = spawn_ls(p[1], p[3])
     usleep(500000)
     v5 = (p[0],p[1],p[2])
     pid[3] = spawn_ls(p[1], p[3])
     usleep(500000)
     v5 = (p[0],p[1],p[2])

@@ -370,29 +370,29 @@ if do_run[6]:
 if do_run[7]:
     print "test7: 4-process lock_server, send signal 2 to first server, kill fourth server, kill other servers, restart other servers, restart fourth server, check logs"
     start_nodes(4,"ls")
 if do_run[7]:
     print "test7: 4-process lock_server, send signal 2 to first server, kill fourth server, kill other servers, restart other servers, restart fourth server, check logs"
     start_nodes(4,"ls")

-    print "Sending paxos breakpoint 2 to first server on port $p[0]"
+    print "Sending paxos breakpoint 2 to first server on port %s" % (p[0],)

     spawn("./rsm_tester", p[0]+1, "breakpoint", 4)
     usleep(300000)
     spawn("./rsm_tester", p[0]+1, "breakpoint", 4)
     usleep(300000)

-    print "Kill fourth server (PID: $pid[3]) on port $p[3]"
+    print "Kill fourth server (PID: %s) on port %s" % (pid[3], p[3])

     os.kill(pid[3], signal.SIGTERM)
     usleep(500000)
     os.kill(pid[3], signal.SIGTERM)
     usleep(500000)

-    print "Kill third server (PID: $pid[2]) on port $p[2]"
+    print "Kill third server (PID: %s) on port %s" % (pid[2], p[2])

     os.kill(pid[2], signal.SIGTERM)
     os.kill(pid[2], signal.SIGTERM)

-    print "Kill second server (PID: $pid[1]) on port $p[1]"
+    print "Kill second server (PID: %s) on port %s" % (pid[1], p[1])

     os.kill(pid[1], signal.SIGTERM)
     usleep(500000)
     os.kill(pid[1], signal.SIGTERM)
     usleep(500000)

-    print "Restarting second server on port $p[1]"
+    print "Restarting second server on port %s" % (p[1],)

     pid[1] = spawn_ls(p[0], p[1])
     usleep(500000)
     pid[1] = spawn_ls(p[0], p[1])
     usleep(500000)

-    print "Restarting third server on port $p[2]"
+    print "Restarting third server on port %s" % (p[2],)

     pid[2] = spawn_ls(p[0], p[2])
     usleep(500000)
     #no view change is possible by now because there is no majority
     for port in (p[1],p[2]):
         num_v = get_num_views(paxos_log(port), port)
         if num_v != in_views[port]:
     pid[2] = spawn_ls(p[0], p[2])
     usleep(500000)
     #no view change is possible by now because there is no majority
     for port in (p[1],p[2]):
         num_v = get_num_views(paxos_log(port), port)
         if num_v != in_views[port]:

-            die("$num_v views in ", paxos_log(port), " : no new views should be formed due to the lack of majority")
-    print "Restarting fourth server on port $p[3]"
+            die("%s_v views in ", paxos_log(port), " : no new views should be formed due to the lack of majority" % (num,))
+    print "Restarting fourth server on port %s" % (p[3],)

     pid[3] = spawn_ls(p[1], p[3])
     usleep(500000)
     v5 = (p[0], p[1], p[2])
     pid[3] = spawn_ls(p[1], p[3])
     usleep(500000)
     v5 = (p[0], p[1], p[2])

@@ -410,11 +410,11 @@ if do_run[7]:
 if do_run[8]:
     print "test8: start 3-process lock service"
     start_nodes(3,"ls")
 if do_run[8]:
     print "test8: start 3-process lock service"
     start_nodes(3,"ls")

-    print "Start lock_tester $p[0]"
+    print "Start lock_tester %s" % (p[0],)

     t = spawn("./lock_tester", p[0])
     t = spawn("./lock_tester", p[0])

-    print "   Wait for lock_tester to finish (waitpid $t)"
+    print "   Wait for lock_tester to finish (waitpid %s)" % (t,)

     waitpid_to(t, 600)
     waitpid_to(t, 600)

-    if os.system("grep \"passed all tests successfully\" lock_tester-$p[0].log"):
+    if os.system("grep \"passed all tests successfully\" lock_tester-%s.log" % (p[0],)):

         mydie("Failed lock tester for test 8")
     cleanup()
     usleep(200000)
         mydie("Failed lock tester for test 8")
     cleanup()
     usleep(200000)

@@ -422,18 +422,18 @@ if do_run[8]:
 if do_run[9]:
     print "test9: start 3-process rsm, kill second slave while lock_tester is running"
     start_nodes(3,"ls")
 if do_run[9]:
     print "test9: start 3-process rsm, kill second slave while lock_tester is running"
     start_nodes(3,"ls")

-    print "Start lock_tester $p[0]"
+    print "Start lock_tester %s" % (p[0],)

     t = spawn("./lock_tester", p[0])
     usleep(random.randint(1,1000000))
     t = spawn("./lock_tester", p[0])
     usleep(random.randint(1,1000000))

-    print "Kill slave (PID: $pid[2]) on port $p[2]"
+    print "Kill slave (PID: %s) on port %s" % (pid[2], p[2])

     os.kill(pid[2], signal.SIGTERM)
     usleep(300000)
     # it should go through 4 views
     v4 = (p[0], p[1])
     wait_and_check_expected_view(v4)
     os.kill(pid[2], signal.SIGTERM)
     usleep(300000)
     # it should go through 4 views
     v4 = (p[0], p[1])
     wait_and_check_expected_view(v4)

-    print "   Wait for lock_tester to finish (waitpid $t)"
+    print "   Wait for lock_tester to finish (waitpid %s)" % (t,)

     waitpid_to(t, 600)
     waitpid_to(t, 600)

-    if os.system("grep \"passed all tests successfully\" lock_tester-$p[0].log"):
+    if os.system("grep \"passed all tests successfully\" lock_tester-%s.log" % (p[0],)):

         mydie("Failed lock tester for test 9")
     cleanup()
     usleep(200000)
         mydie("Failed lock tester for test 9")
     cleanup()
     usleep(200000)

@@ -441,23 +441,23 @@ if do_run[9]:
 if do_run[10]:
     print "test10: start 3-process rsm, kill second slave and restarts it later while lock_tester is running"
     start_nodes(3,"ls")
 if do_run[10]:
     print "test10: start 3-process rsm, kill second slave and restarts it later while lock_tester is running"
     start_nodes(3,"ls")

-    print "Start lock_tester $p[0]"
+    print "Start lock_tester %s" % (p[0],)

     t = spawn("./lock_tester", p[0])
     usleep(random.randint(1,1000000))
     t = spawn("./lock_tester", p[0])
     usleep(random.randint(1,1000000))

-    print "Kill slave (PID: $pid[2]) on port $p[2]"
+    print "Kill slave (PID: %s) on port %s" % (pid[2], p[2])

     os.kill(pid[2], signal.SIGTERM)
     usleep(300000)
     # it should go through 4 views
     v4 = (p[0], p[1])
     wait_and_check_expected_view(v4)
     usleep(300000)
     os.kill(pid[2], signal.SIGTERM)
     usleep(300000)
     # it should go through 4 views
     v4 = (p[0], p[1])
     wait_and_check_expected_view(v4)
     usleep(300000)

-    print "Restarting killed lock_server on port $p[2]"
+    print "Restarting killed lock_server on port %s" % (p[2],)

     pid[2] = spawn_ls(p[0], p[2])
     v5 = (p[0],p[1],p[2])
     wait_and_check_expected_view(v5)
     pid[2] = spawn_ls(p[0], p[2])
     v5 = (p[0],p[1],p[2])
     wait_and_check_expected_view(v5)

-    print "   Wait for lock_tester to finish (waitpid $t)"
+    print "   Wait for lock_tester to finish (waitpid %s)" % (t,)

     waitpid_to(t, 600)
     waitpid_to(t, 600)

-    if os.system("grep \"passed all tests successfully\" lock_tester-$p[0].log"):
+    if os.system("grep \"passed all tests successfully\" lock_tester-%s.log" % (p[0],)):

         mydie("Failed lock tester for test 10")
     cleanup()
     usleep(200000)
         mydie("Failed lock tester for test 10")
     cleanup()
     usleep(200000)

@@ -465,18 +465,18 @@ if do_run[10]:
 if do_run[11]:
     print "test11: start 3-process rsm, kill primary while lock_tester is running"
     start_nodes(3,"ls")
 if do_run[11]:
     print "test11: start 3-process rsm, kill primary while lock_tester is running"
     start_nodes(3,"ls")

-    print "Start lock_tester $p[0]"
+    print "Start lock_tester %s" % (p[0],)

     t = spawn("./lock_tester", p[0])
     usleep(random.randint(1,1000000))
     t = spawn("./lock_tester", p[0])
     usleep(random.randint(1,1000000))

-    print "Kill primary (PID: $pid[0]) on port $p[0]"
+    print "Kill primary (PID: %s) on port %s" % (pid[0], p[0])

     os.kill(pid[0], signal.SIGTERM)
     usleep(300000)
     # it should go through 4 views
     v4 = (p[1], p[2])
     wait_and_check_expected_view(v4)
     os.kill(pid[0], signal.SIGTERM)
     usleep(300000)
     # it should go through 4 views
     v4 = (p[1], p[2])
     wait_and_check_expected_view(v4)

-    print "   Wait for lock_tester to finish (waitpid $t)"
+    print "   Wait for lock_tester to finish (waitpid %s)" % (t,)

     waitpid_to(t, 600)
     waitpid_to(t, 600)

-    if os.system("grep \"passed all tests successfully\" lock_tester-$p[0].log"):
+    if os.system("grep \"passed all tests successfully\" lock_tester-%s.log" % (p[0],)):

         mydie("Failed lock tester for test 11")
     cleanup()
     usleep(200000)
         mydie("Failed lock tester for test 11")
     cleanup()
     usleep(200000)

@@ -484,16 +484,16 @@ if do_run[11]:
 if do_run[12]:
     print "test12: start 3-process rsm, kill master at break1 and restart it while lock_tester is running"
     start_nodes(3, "ls")
 if do_run[12]:
     print "test12: start 3-process rsm, kill master at break1 and restart it while lock_tester is running"
     start_nodes(3, "ls")

-    print "Start lock_tester $p[0]"
+    print "Start lock_tester %s" % (p[0],)

     t = spawn("./lock_tester", p[0])
     usleep(100000)
     t = spawn("./lock_tester", p[0])
     usleep(100000)

-    print "Kill master (PID: $pid[0]) on port $p[0] at breakpoint 1"
+    print "Kill master (PID: %s) on port %s at breakpoint 1" % (pid[0], p[0])

     spawn("./rsm_tester", p[0]+1, "breakpoint", 1)
     usleep(100000)
     # it should go through 5 views
     v4 = (p[1], p[2])
     wait_and_check_expected_view(v4)
     spawn("./rsm_tester", p[0]+1, "breakpoint", 1)
     usleep(100000)
     # it should go through 5 views
     v4 = (p[1], p[2])
     wait_and_check_expected_view(v4)

-    print "Restarting killed lock_server on port $p[0]"
+    print "Restarting killed lock_server on port %s" % (p[0],)

     pid[0] = spawn_ls(p[1], p[0])
     usleep(300000)
     # the last view should include all nodes
     pid[0] = spawn_ls(p[1], p[0])
     usleep(300000)
     # the last view should include all nodes

@@ -502,9 +502,9 @@ if do_run[12]:
         wait_for_view_change(paxos_log(port), in_views[port]+1, port, 20)
     for port in lastv:
         check_views(paxos_log(port), views, lastv)
         wait_for_view_change(paxos_log(port), in_views[port]+1, port, 20)
     for port in lastv:
         check_views(paxos_log(port), views, lastv)

-    print "   Wait for lock_tester to finish (waitpid $t)"
+    print "   Wait for lock_tester to finish (waitpid %s)" % (t,)

     waitpid_to(t, 600)
     waitpid_to(t, 600)

-    if os.system("grep \"passed all tests successfully\" lock_tester-$p[0].log"):
+    if os.system("grep \"passed all tests successfully\" lock_tester-%s.log" % (p[0],)):

         mydie("Failed lock tester for test 12")
     cleanup()
     usleep(200000)
         mydie("Failed lock tester for test 12")
     cleanup()
     usleep(200000)

@@ -512,16 +512,16 @@ if do_run[12]:
 if do_run[13]:
     print "test13: start 3-process rsm, kill slave at break1 and restart it while lock_tester is running"
     start_nodes(3, "ls")
 if do_run[13]:
     print "test13: start 3-process rsm, kill slave at break1 and restart it while lock_tester is running"
     start_nodes(3, "ls")

-    print "Start lock_tester $p[0]"
+    print "Start lock_tester %s" % (p[0],)

     t = spawn("./lock_tester", p[0])
     usleep(100000)
     t = spawn("./lock_tester", p[0])
     usleep(100000)

-    print "Kill slave (PID: $pid[2]) on port $p[2] at breakpoint 1"
+    print "Kill slave (PID: %s) on port %s at breakpoint 1" % (pid[2], p[2])

     spawn("./rsm_tester", p[2]+1, "breakpoint", 1)
     usleep(100000)
     # it should go through 4 views
     v4 = (p[0], p[1])
     wait_and_check_expected_view(v4)
     spawn("./rsm_tester", p[2]+1, "breakpoint", 1)
     usleep(100000)
     # it should go through 4 views
     v4 = (p[0], p[1])
     wait_and_check_expected_view(v4)

-    print "Restarting killed lock_server on port $p[2]"
+    print "Restarting killed lock_server on port %s" % (p[2],)

     pid[2] = spawn_ls(p[0], p[2])
     usleep(300000)
     # the last view should include all nodes
     pid[2] = spawn_ls(p[0], p[2])
     usleep(300000)
     # the last view should include all nodes

@@ -530,9 +530,9 @@ if do_run[13]:
         wait_for_view_change(paxos_log(port), in_views[port]+1, port, 20)
     for port in lastv:
         check_views(paxos_log(port), views, lastv)
         wait_for_view_change(paxos_log(port), in_views[port]+1, port, 20)
     for port in lastv:
         check_views(paxos_log(port), views, lastv)

-    print "   Wait for lock_tester to finish (waitpid $t)"
+    print "   Wait for lock_tester to finish (waitpid %s)" % (t,)

     waitpid_to(t, 600)
     waitpid_to(t, 600)

-    if os.system("grep \"passed all tests successfully\" lock_tester-$p[0].log"):
+    if os.system("grep \"passed all tests successfully\" lock_tester-%s.log" % (p[0],)):

         mydie("Failed lock tester for test 13")
     cleanup()
     usleep(200000)
         mydie("Failed lock tester for test 13")
     cleanup()
     usleep(200000)

@@ -540,12 +540,12 @@ if do_run[13]:
 if do_run[14]:
     print "test14: start 5-process rsm, kill slave break1, kill slave break2"
     start_nodes(5, "ls")
 if do_run[14]:
     print "test14: start 5-process rsm, kill slave break1, kill slave break2"
     start_nodes(5, "ls")

-    print "Start lock_tester $p[0]"
+    print "Start lock_tester %s" % (p[0],)

     t = spawn("./lock_tester", p[0])
     usleep(100000)
     t = spawn("./lock_tester", p[0])
     usleep(100000)

-    print "Kill slave (PID: $pid[4]) on port $p[4] at breakpoint 1"
+    print "Kill slave (PID: %s) on port %s at breakpoint 1" % (pid[4], p[4])

     spawn("./rsm_tester", p[4]+1, "breakpoint", 1)
     spawn("./rsm_tester", p[4]+1, "breakpoint", 1)

-    print "Kill slave (PID: $pid[3]) on port $p[3] at breakpoint 2"
+    print "Kill slave (PID: %s) on port %s at breakpoint 2" % (pid[3], p[3])

     spawn("./rsm_tester", p[3]+1, "breakpoint", 2)
     usleep(100000)
     # two view changes:
     spawn("./rsm_tester", p[3]+1, "breakpoint", 2)
     usleep(100000)
     # two view changes:

@@ -557,9 +557,9 @@ if do_run[14]:
     lastv = (p[0],p[1],p[2])
     for port in lastv:
         wait_for_view_change(paxos_log(port), in_views[port]+1, port, 20)
     lastv = (p[0],p[1],p[2])
     for port in lastv:
         wait_for_view_change(paxos_log(port), in_views[port]+1, port, 20)

-    print "   Wait for lock_tester to finish (waitpid $t)"
+    print "   Wait for lock_tester to finish (waitpid %s)" % (t,)

     waitpid_to(t, 600)
     waitpid_to(t, 600)

-    if os.system("grep \"passed all tests successfully\" lock_tester-$p[0].log"):
+    if os.system("grep \"passed all tests successfully\" lock_tester-%s.log" % (p[0],)):

         mydie("Failed lock tester for test 14")
     cleanup()
     usleep(200000)
         mydie("Failed lock tester for test 14")
     cleanup()
     usleep(200000)

@@ -567,12 +567,12 @@ if do_run[14]:
 if do_run[15]:
     print "test15: start 5-process rsm, kill slave break1, kill primary break2"
     start_nodes(5, "ls")
 if do_run[15]:
     print "test15: start 5-process rsm, kill slave break1, kill primary break2"
     start_nodes(5, "ls")

-    print "Start lock_tester $p[0]"
+    print "Start lock_tester %s" % (p[0],)

     t = spawn("./lock_tester", p[0])
     usleep(100000)
     t = spawn("./lock_tester", p[0])
     usleep(100000)

-    print "Kill slave (PID: $pid[4]) on port $p[4] at breakpoint 1"
+    print "Kill slave (PID: %s) on port %s at breakpoint 1" % (pid[4], p[4])

     spawn("./rsm_tester", p[4]+1, "breakpoint", 1)
     spawn("./rsm_tester", p[4]+1, "breakpoint", 1)

-    print "Kill primary (PID: $pid[0]) on port $p[0] at breakpoint 2"
+    print "Kill primary (PID: %s) on port %s at breakpoint 2" % (pid[0], p[0])

     spawn("./rsm_tester", p[0]+1, "breakpoint", 2)
     usleep(100000)
     # two view changes:
     spawn("./rsm_tester", p[0]+1, "breakpoint", 2)
     usleep(100000)
     # two view changes:

@@ -584,9 +584,9 @@ if do_run[15]:
     lastv = (p[1],p[2],p[3])
     for port in lastv:
         wait_for_view_change(paxos_log(port), in_views[port]+1, port, 20)
     lastv = (p[1],p[2],p[3])
     for port in lastv:
         wait_for_view_change(paxos_log(port), in_views[port]+1, port, 20)

-    print "   Wait for lock_tester to finish (waitpid $t)"
+    print "   Wait for lock_tester to finish (waitpid %s)" % (t,)

     waitpid_to(t, 600)
     waitpid_to(t, 600)

-    if os.system("grep \"passed all tests successfully\" lock_tester-$p[0].log"):
+    if os.system("grep \"passed all tests successfully\" lock_tester-%s.log" % (p[0],)):

         mydie("Failed lock tester for test 15")
     cleanup()
     usleep(200000)
         mydie("Failed lock tester for test 15")
     cleanup()
     usleep(200000)

@@ -594,10 +594,10 @@ if do_run[15]:
 if do_run[16]:
     print "test16: start 3-process rsm, partition primary, heal it"
     start_nodes(3, "ls")
 if do_run[16]:
     print "test16: start 3-process rsm, partition primary, heal it"
     start_nodes(3, "ls")

-    print "Start lock_tester $p[0]"
+    print "Start lock_tester %s" % (p[0],)

     t = spawn("./lock_tester", p[0])
     usleep(100000)
     t = spawn("./lock_tester", p[0])
     usleep(100000)

-    print "Partition primary (PID: $pid[0]) on port $p[0] at breakpoint"
+    print "Partition primary (PID: %s) on port %s at breakpoint" % (pid[0], p[0])

     spawn("./rsm_tester", p[0]+1, "partition", 0)
     usleep(300000)
     print "first view change wait"
     spawn("./rsm_tester", p[0]+1, "partition", 0)
     usleep(300000)
     print "first view change wait"

@@ -605,7 +605,7 @@ if do_run[16]:
     for port in lastv:
         wait_for_view_change(paxos_log(port), in_views[port]+1, port, 20)
     usleep(100000)
     for port in lastv:
         wait_for_view_change(paxos_log(port), in_views[port]+1, port, 20)
     usleep(100000)

-    print "Heal partition primary (PID: $pid[0]) on port $p[0] at breakpoint"
+    print "Heal partition primary (PID: %s) on port %s at breakpoint" % (pid[0], p[0])

     spawn("./rsm_tester", p[0]+1, "partition", 1)
     usleep(100000)
     # xxx it should test that this is the 5th view!
     spawn("./rsm_tester", p[0]+1, "partition", 1)
     usleep(100000)
     # xxx it should test that this is the 5th view!

@@ -613,9 +613,9 @@ if do_run[16]:
     lastv = (p[0], p[1],p[2])
     for port in lastv:
         wait_for_view_change(paxos_log(port), in_views[port]+1, port, 20)
     lastv = (p[0], p[1],p[2])
     for port in lastv:
         wait_for_view_change(paxos_log(port), in_views[port]+1, port, 20)

-    print "   Wait for lock_tester to finish (waitpid $t)"
+    print "   Wait for lock_tester to finish (waitpid %s)" % (t,)

     waitpid_to(t, 600)
     waitpid_to(t, 600)

-    if os.system("grep \"passed all tests successfully\" lock_tester-$p[0].log"):
+    if os.system("grep \"passed all tests successfully\" lock_tester-%s.log" % (p[0],)):

         mydie("Failed lock tester for test 16")
     cleanup()
     usleep(200000)
         mydie("Failed lock tester for test 16")
     cleanup()
     usleep(200000)