- client and API: improve the way an app checks for the death of the …

…client

    Old: heartbeat mechanism
    Problem: if the client is blocked for > 30 secs
        (e.g. because it takes a long time to write the state file,
        of because it's stopped in a debugger)
        then apps exit.
        This is bad is the app doesn't checkpoint and has been
        running for a long time.
    New: the client passes its PID to the app.
        The app periodically (10 sec) checks that the process still exists.
    Notes:
    - For backward compatibility (e.g. new API w/ old client,
        or vice versa) the client still sends heartbeats,
        and the API checks heartbeats if the client doesn't pass a PID.
    - The new mechanism works only if the client's PID isn't assigned
        to a new process within 10 secs of the client exiting.
        Windows 2000 reuses PIDs immediately, so check for Win2K
        and don't use this mechanism if so.

TODO: For Unix multithread apps,
    critical sections aren't currently being enforced.
    Need to fix this by masking signals.

svn path=/trunk/boinc/; revision=26147

checkin_notes

@@ -4482,6 +4482,37 @@ David  21 Sept 2012
     clientgui/
         MainDocument.cpp
 
+David  11 Oct 2012
+    - client and API: improve the way an app checks for the death of the client
+        Old: heartbeat mechanism
+        Problem: if the client is blocked for > 30 secs
+            (e.g. because it takes a long time to write the state file,
+            of because it's stopped in a debugger)
+            then apps exit.
+            This is bad is the app doesn't checkpoint and has been
+            running for a long time.
+        New: the client passes its PID to the app.
+            The app periodically (10 sec) checks that the process still exists.
+        Notes:
+        - For backward compatibility (e.g. new API w/ old client,
+            or vice versa) the client still sends heartbeats,
+            and the API checks heartbeats if the client doesn't pass a PID.
+        - The new mechanism works only if the client's PID isn't assigned
+            to a new process within 10 secs of the client exiting.
+            Windows 2000 reuses PIDs immediately, so check for Win2K
+            and don't use this mechanism if so.
+
+    TODO: For Unix multithread apps,
+        critical sections aren't currently being enforced.
+        Need to fix this by masking signals.
+
+    client/
+        hostinfo_win.cpp
+        app_start.cpp
+    lib/
+        app_ipc.cpp,h
+        proc_control.cpp
+
 Charlie 15 Oct 2012
     - MGR: We don't save Simple View's width & height since it's 
         window is not resizable, so don't try to read them back.

client/app_start.cpp

@@ -218,6 +218,17 @@ void ACTIVE_TASK::init_app_init_data(APP_INIT_DATA& aid) {
     relative_to_absolute("", aid.boinc_dir);
     strcpy(aid.authenticator, wup->project->authenticator);
     aid.slot = slot;
+#ifdef _WIN32
+    if (strstr(gstate.hostinfo.os_name, "Windows 2000")) {
+        // Win2K immediately reuses PIDs, so can't use this mechanism
+        //
+        aid.client_pid = 0;
+    } else {
+        aid.client_pid = GetCurrentProcessId();
+    }
+#else
+    aid.client_pid = getpid();
+#endif
     strcpy(aid.wu_name, wup->name);
     strcpy(aid.result_name, result->name);
     aid.user_total_credit = wup->project->user_total_credit;

client/hostinfo_win.cpp

@@ -478,8 +478,7 @@ int get_os_information(
                 strcat(os_name, "Windows 2000");
             }
 
-            if ( osvi.dwMajorVersion <= 4 )
-            {
+            if ( osvi.dwMajorVersion <= 4 ) {
                 strcat(os_name, "Windows NT");
             }
 

lib/app_ipc.cpp

@@ -100,6 +100,7 @@ void APP_INIT_DATA::copy(const APP_INIT_DATA& a) {
     teamid                        = a.teamid;
     hostid                        = a.hostid;
     slot                          = a.slot;
+    client_pid                    = a.client_pid;
     user_total_credit             = a.user_total_credit;
     user_expavg_credit            = a.user_expavg_credit;
     host_total_credit             = a.host_total_credit;
@@ -192,6 +193,7 @@ int write_init_data_file(FILE* f, APP_INIT_DATA& ai) {
 #endif
     fprintf(f,
         "<slot>%d</slot>\n"
+        "<client_pid>%d</client_pid>\n"
         "<wu_cpu_time>%f</wu_cpu_time>\n"
         "<starting_elapsed_time>%f</starting_elapsed_time>\n"
         "<using_sandbox>%d</using_sandbox>\n"
@@ -213,6 +215,7 @@ int write_init_data_file(FILE* f, APP_INIT_DATA& ai) {
         "<rsc_disk_bound>%f</rsc_disk_bound>\n"
         "<computation_deadline>%f</computation_deadline>\n",
         ai.slot,
+        ai.client_pid,
         ai.wu_cpu_time,
         ai.starting_elapsed_time,
         ai.using_sandbox?1:0,
@@ -263,6 +266,7 @@ void APP_INIT_DATA::clear() {
     strcpy(result_name, "");
     strcpy(authenticator, "");
     slot = 0;
+    client_pid = 0;
     user_total_credit = 0;
     user_expavg_credit = 0;
     host_total_credit = 0;
@@ -367,6 +371,7 @@ int parse_init_data_file(FILE* f, APP_INIT_DATA& ai) {
         if (xp.parse_int("shm_key", ai.shmem_seg_name)) continue;
 #endif
         if (xp.parse_int("slot", ai.slot)) continue;
+        if (xp.parse_int("client_pid", ai.client_pid)) continue;
         if (xp.parse_double("user_total_credit", ai.user_total_credit)) continue;
         if (xp.parse_double("user_expavg_credit", ai.user_expavg_credit)) continue;
         if (xp.parse_double("host_total_credit", ai.host_total_credit)) continue;

lib/app_ipc.h

@@ -167,6 +167,7 @@ struct APP_INIT_DATA {
     char result_name[256];
     char authenticator[256];
     int slot;
+    int client_pid;
     double user_total_credit;
     double user_expavg_credit;
     double host_total_credit;

lib/proc_control.cpp

@@ -166,7 +166,8 @@ void kill_descendants() {
     kill_all(descendants);
 }
 #else
-// Same, but if child_pid is nonzero, give it a chance to exit gracefully on Unix
+// Same, but if child_pid is nonzero,
+// give it a chance to exit gracefully on Unix
 //
 void kill_descendants(int child_pid) {
     vector<int> descendants;