Miscellaneous show_help updates

.gitignore

@@ -629,6 +629,7 @@ test/datatype/ddt_raw
 test/datatype/opal_datatype_test
 test/datatype/position_noncontig
 test/datatype/unpack_ooo
+test/datatype/unpack_hetero
 
 test/dss/dss_buffer
 test/dss/dss_copy

opal/mca/btl/tcp/btl_tcp_component.c

@@ -10,7 +10,7 @@
  *                         University of Stuttgart.  All rights reserved.
  * Copyright (c) 2004-2005 The Regents of the University of California.
  *                         All rights reserved.
- * Copyright (c) 2007-2015 Cisco Systems, Inc.  All rights reserved.
+ * Copyright (c) 2007-2017 Cisco Systems, Inc.  All rights reserved
  * Copyright (c) 2008      Sun Microsystems, Inc.  All rights reserved.
  * Copyright (c) 2009      Oak Ridge National Laboratory
  * Copyright (c) 2012-2015 Los Alamos National Security, LLC.  All rights
@@ -1334,6 +1334,34 @@ static void mca_btl_tcp_component_accept_handler( int incoming_sd,
 }
 
 
+static const char *get_peer_name(int fd)
+{
+    char *str;
+    const char *ret;
+    struct sockaddr sa;
+    struct sockaddr_in *si;
+    socklen_t slt = (socklen_t) sizeof(sa);
+
+    int rc = getpeername(fd, &sa, &slt);
+    if (0 != rc) {
+        ret = strdup("Unknown");
+        return ret;
+    }
+
+    str = malloc(INET_ADDRSTRLEN);
+    if (NULL == str) {
+        return NULL;
+    }
+
+    si = (struct sockaddr_in*) &sa;
+    ret = inet_ntop(AF_INET, &(si->sin_addr), str, INET_ADDRSTRLEN);
+    if (NULL == ret) {
+        free(str);
+    }
+
+    return ret;
+}
+
 /**
  * Event callback when there is data available on the registered
  * socket to recv. This callback is triggered only once per lifetime
@@ -1365,20 +1393,22 @@ static void mca_btl_tcp_component_recv_handler(int sd, short flags, void* user)
         if (ENOPROTOOPT == errno) {
             sockopt = false;
         } else {
-            opal_output_verbose(20, opal_btl_base_framework.framework_output,
-                                "Cannot get current recv timeout value of the socket"
-                                "Local_host:%s PID:%d",
-                                opal_process_info.nodename, getpid());
+            opal_show_help("help-mpi-btl-tcp.txt", "socket flag fail",
+                           true, opal_process_info.nodename,
+                           getpid(),
+                           "getsockopt(sd, SOL_SOCKET, SO_RCVTIMEO, ...)",
+                           strerror(opal_socket_errno), opal_socket_errno);
             return;
         }
     } else {
         tv.tv_sec = 2;
         tv.tv_usec = 0;
         if (0 != setsockopt(sd, SOL_SOCKET, SO_RCVTIMEO, &tv, sizeof(tv))) {
-            opal_output_verbose(20, opal_btl_base_framework.framework_output,
-                                "Cannot set new recv timeout value of the socket"
-                                "Local_host:%s PID:%d",
-                                opal_process_info.nodename, getpid());
+            opal_show_help("help-mpi-btl-tcp.txt", "socket flag fail",
+                           true, opal_process_info.nodename,
+                           getpid(),
+                           "setsockopt(sd, SOL_SOCKET, SO_RCVTIMEO, ...)",
+                           strerror(opal_socket_errno), opal_socket_errno);
            return;
        }
     }
@@ -1397,14 +1427,16 @@ static void mca_btl_tcp_component_recv_handler(int sd, short flags, void* user)
      * This attempted connection will be ignored; your MPI job may or may not
      * continue properly.
      */
-     if (sizeof(hs_msg) != retval) {
-         opal_output_verbose(20, opal_btl_base_framework.framework_output,
-                            "process did not receive full connect ACK "
-                            "Local_host:%s PID:%d String_received:%s Test_fail:%s",
-                            opal_process_info.nodename,
-                            getpid(),
-                            (retval > 0) ? hs_msg.magic_id : "<nothing>",
-                            "handshake message length");
+    if (sizeof(hs_msg) != retval) {
+         const char *peer = get_peer_name(sd);
+         opal_show_help("help-mpi-btl-tcp.txt",
+                        "did not receive full magic id string",
+                        true,
+                        opal_process_info.nodename,
+                        getpid(),
+                        opal_version_string,
+                        peer);
+         free((char*) peer);
 
          /* The other side probably isn't OMPI, so just hang up */
          CLOSE_THE_SOCKET(sd);
@@ -1413,12 +1445,18 @@ static void mca_btl_tcp_component_recv_handler(int sd, short flags, void* user)
 
     guid = hs_msg.guid;
     if (0 != strncmp(hs_msg.magic_id, mca_btl_tcp_magic_id_string, len)) {
-        opal_output_verbose(20, opal_btl_base_framework.framework_output,
-                            "process did not receive right magic string. "
-                            "Local_host:%s PID:%d String_received:%s Test_fail:%s",
-                            opal_process_info.nodename,
-                            getpid(), hs_msg.magic_id,
-                            "string value");
+         const char *peer = get_peer_name(sd);
+         opal_show_help("help-mpi-btl-tcp.txt",
+                        "received incorrect magic id string",
+                        true,
+                        opal_process_info.nodename,
+                        getpid(),
+                        opal_version_string,
+                        peer,
+                        hs_msg.magic_id,
+                        mca_btl_tcp_magic_id_string);
+         free((char*) peer);
+
         /* The other side probably isn't OMPI, so just hang up */
         CLOSE_THE_SOCKET(sd);
         return;
@@ -1427,10 +1465,11 @@ static void mca_btl_tcp_component_recv_handler(int sd, short flags, void* user)
     if (sockopt) {
        /* reset RECVTIMEO option to its original state */
        if (0 != setsockopt(sd, SOL_SOCKET, SO_RCVTIMEO, &save, sizeof(save))) {
-           opal_output_verbose(20, opal_btl_base_framework.framework_output,
-                               "Cannot reset recv timeout value"
-                               "Local_host:%s PID:%d",
-                               opal_process_info.nodename, getpid());
+           opal_show_help("help-mpi-btl-tcp.txt", "socket flag fail",
+                          true, opal_process_info.nodename,
+                          getpid(),
+                          "setsockopt(sd, SOL_SOCKET, SO_RCVTIMEO, ...)",
+                          strerror(opal_socket_errno), opal_socket_errno);
           return;
        }
     }

opal/mca/btl/tcp/help-mpi-btl-tcp.txt

@@ -35,7 +35,7 @@ values are in the range [1 .. 2^16-1]. This value will be ignored
 WARNING: Open MPI failed to TCP connect to a peer MPI process.  This
 should not happen.
 
-Your Open MPI job may now fail.
+Your Open MPI job may now hang or fail.
 
   Local host: %s
   PID:        %d
@@ -46,7 +46,7 @@ Your Open MPI job may now fail.
 WARNING: Open MPI failed to handshake with a connecting peer MPI
 process over TCP.  This should not happen.
 
-Your Open MPI job may now fail.
+Your Open MPI job may now hang or fail.
 
   Local host: %s
   PID:        %d
@@ -102,8 +102,11 @@ hopefully be able to continue).
   Known IPs of peer:   %s
 #
 [socket flag fail]
-WARNING: Open MPI failed to set flags on a TCP socket.  This should
-not happen.  It is likely that your MPI job will now fail.
+WARNING: Open MPI failed to get or set flags on a TCP socket.  This
+should not happen.
+
+This may cause unpredictable behavior, and may end up hanging or
+aborting your job.
 
   Local host: %s
   PID:        %d
@@ -164,4 +167,43 @@ Your Open MPI job may now fail.
   PID:        %d
   Message:    %s
   Error:      %s (%d)
-#
+#
+[did not receive full magic id string]
+The TCP BTL received an inbound socket connection from an unidentified
+peer.  This typically means one of two things:
+
+1. A non-Open MPI process tried to connect to this Open MPI process.
+2. An Open MPI process compiled against a different version of Open
+   MPI tried to connect to this Open MPI process.
+
+Open MPI only supports running exactly the same version between all
+processes in a single job.
+
+This may cause unpredictable behavior, and may end up aborting your
+job.
+
+  Local host:             %s
+  Local PID:              %d
+  Local Open MPI version: %s
+  Peer IP address:        %s
+#
+[received incorrect magic id string]
+The TCP BTL received an inbound socket connection from a peer that did
+not identify itself correctly as an Open MPI process. This typically
+means one of two things:
+
+1. A non-Open MPI process tried to connect to this Open MPI process.
+2. An Open MPI process compiled against a different version of Open
+   MPI tried to connect to this Open MPI process.
+
+Open MPI only supports running exactly the same version between all
+processes in a single job.
+
+This may cause unpredictable behavior, and may end up hanging or
+aborting your job.
+
+  Local host:             %s
+  Local PID:              %d
+  Local Open MPI version: %s
+  Peer IP address:        %s
+  Peer identifier:        %s (expected %s)

opal/util/net.c

@@ -11,7 +11,7 @@
  *                         All rights reserved.
  * Copyright (c) 2007      Los Alamos National Security, LLC.  All rights
  *                         reserved.
- * Copyright (c) 2009-2015 Cisco Systems, Inc.  All rights reserved.
+ * Copyright (c) 2009-2017 Cisco Systems, Inc.  All rights reserved
  * Copyright (c) 2013      Intel, Inc.  All rights reserved.
  * Copyright (c) 2015      Research Organization for Information Science
  *                         and Technology (RIST). All rights reserved.
@@ -361,7 +361,9 @@ opal_net_addr_isipv4public(const struct sockaddr *addr)
 bool
 opal_net_addr_isipv6linklocal(const struct sockaddr *addr)
 {
+#if OPAL_ENABLE_IPV6
     struct sockaddr_in6 if_addr;
+#endif
 
     switch (addr->sa_family) {
 #if OPAL_ENABLE_IPV6

opal/util/show_help.c

@@ -9,7 +9,7 @@
  *                         University of Stuttgart.  All rights reserved.
  * Copyright (c) 2004-2005 The Regents of the University of California.
  *                         All rights reserved.
- * Copyright (c) 2008      Cisco Systems, Inc.  All rights reserved.
+ * Copyright (c) 2008-2017 Cisco Systems, Inc.  All rights reserved
  * Copyright (c) 2015      Research Organization for Information Science
  *                         and Technology (RIST). All rights reserved.
  * $COPYRIGHT$
@@ -36,24 +36,22 @@
 #include "opal/constants.h"
 
 
+/*
+ * Local functions
+ */
+static int default_backend(const char *filename, const char *topic,
+                           bool want_error_header, va_list arglist);
+
+
 /*
  * Private variables
  */
 static const char *default_filename = "help-messages";
 static const char *dash_line = "--------------------------------------------------------------------------\n";
 static int output_stream = -1;
 static char **search_dirs = NULL;
-
-/*
- * Local functions
- */
-static int opal_show_vhelp_internal(const char *filename, const char *topic,
-                                    bool want_error_header, va_list arglist);
-static int opal_show_help_internal(const char *filename, const char *topic,
-                                   bool want_error_header, ...);
-
-opal_show_help_fn_t opal_show_help = opal_show_help_internal;
-opal_show_vhelp_fn_t opal_show_vhelp = opal_show_vhelp_internal;
+static volatile bool enabled = true;
+static opal_show_help_internal_fn_t registered_backend = default_backend;
 
 
 int opal_show_help_init(void)
@@ -330,8 +328,17 @@ char *opal_show_help_string(const char *filename, const char *topic,
     return output;
 }
 
-static int opal_show_vhelp_internal(const char *filename, const char *topic,
-                                    bool want_error_header, va_list arglist)
+/* This is the default back-end to all the opal_show_*help()
+ * functions.  All it does it render the string and then call
+ * opal_output().
+ *
+ * All decisions about whether show_help messages are enabled or not
+ * have been made by the time this function is invoked.
+ */
+static int default_backend(const char *filename,
+                           const char *topic,
+                           bool want_error_header,
+                           va_list arglist)
 {
     char *output;
 
@@ -348,20 +355,68 @@ static int opal_show_vhelp_internal(const char *filename, const char *topic,
     return (NULL == output) ? OPAL_ERROR : OPAL_SUCCESS;
 }
 
-static int opal_show_help_internal(const char *filename, const char *topic,
-                                   bool want_error_header, ...)
+int opal_show_vhelp(const char *filename, const char *topic,
+                    bool want_error_header, va_list arglist)
+{
+    if (!enabled) {
+        return OPAL_SUCCESS;
+    }
+
+    return registered_backend(filename, topic,
+                              want_error_header, arglist);
+}
+
+int opal_show_vhelp_final(const char *filename, const char *topic,
+                          bool want_error_header, va_list arglist)
+{
+    if (!enabled) {
+        return OPAL_SUCCESS;
+    }
+
+    enabled = false;
+
+    return registered_backend(filename, topic,
+                              want_error_header, arglist);
+}
+
+int opal_show_help(const char *filename, const char *topic,
+                   bool want_error_header, ...)
 {
     va_list arglist;
-    int rc;
+    int rc = OPAL_SUCCESS;
 
-    /* Convert it to a single string */
     va_start(arglist, want_error_header);
     rc = opal_show_vhelp(filename, topic, want_error_header, arglist);
     va_end(arglist);
 
     return rc;
 }
 
+int opal_show_help_final(const char *filename, const char *topic,
+                                bool want_error_header, ...)
+{
+    va_list arglist;
+    int rc = OPAL_SUCCESS;
+
+    va_start(arglist, want_error_header);
+    rc = opal_show_vhelp_final(filename, topic, want_error_header,
+                               arglist);
+    va_end(arglist);
+
+    return rc;
+}
+
+int opal_show_help_register_backend(opal_show_help_internal_fn_t func)
+{
+    if (NULL == func) {
+        registered_backend = default_backend;
+    } else {
+        registered_backend = func;
+    }
+
+    return OPAL_SUCCESS;
+}
+
 int opal_show_help_add_dir(const char *directory)
 {
     opal_argv_append_nosize(&search_dirs, directory);