Changeset 37792

Timestamp:

Jan 11, 2015, 1:08:36 PM (12 years ago)

Author:

eugene

Message:

harvest unrelated children and do not exit

Location:

branches/eam_branches/ipp-20140904/Ohana/src/libdvo/src

Files:

• HostTable.c (modified) (5 diffs)
• dvo_convert.c (modified) (1 diff)

branches/eam_branches/ipp-20140904/Ohana/src/libdvo/src/HostTable.c

@@ -333 +333 @@
 
   // this loop has 2 chunks: (a) check for I/O + (b) check for jobs done
-  while (1) {
+  while (Nfound < Nrunning) {
 
     // Wait up to 0.5 second for host to provide I/O
@@ -362 +362 @@
         // if (table->hosts[i].stdout.Nbuffer > 0x10000) {
         if (table->hosts[i].stdout.Nbuffer > 0x1000) {
-          int printHead = VERBOSE || (table->hosts[i].stdout.Nbuffer > 0);
+          int printHead = VERBOSE && (table->hosts[i].stdout.Nbuffer > 0);
           if (printHead) fprintf (stdout, "--- stdout from %s --- (%d bytes, v1)\n", table->hosts[i].hostname, table->hosts[i].stdout.Nbuffer);
           int Nout = write (STDOUT_FILENO, table->hosts[i].stdout.buffer, table->hosts[i].stdout.Nbuffer);
@@ -376 +376 @@
         // if (table->hosts[i].stderr.Nbuffer > 0x10000) {
         if (table->hosts[i].stderr.Nbuffer > 0x1000) {
-          int printHead = VERBOSE || (table->hosts[i].stderr.Nbuffer > 0);
+          int printHead = VERBOSE && (table->hosts[i].stderr.Nbuffer > 0);
           if (printHead) fprintf (stdout, "--- stderr from %s --- (%d bytes, v1)\n", table->hosts[i].hostname, table->hosts[i].stderr.Nbuffer);
           int Nout = write (STDOUT_FILENO, table->hosts[i].stderr.buffer, table->hosts[i].stderr.Nbuffer);
@@ -392 +392 @@
       if (!pid) {
         // fprintf (stderr, "no hosts to harvest\n");
-        usleep (500000);
+        usleep (200000);
         break; // no outstanding jobs have finished 
       }
@@ -445 +445 @@
         }
       }
-      if (!found) {
-        fprintf (stderr, "Programming error: failed to matched finished job to known host!\n");
-        exit (2);
-      }
-      Nfound ++;
+      if (found) Nfound ++;
+      // we sometimes harvest children not in the list (eg, closed kapa window).  just ignore
       if (Nfound == Nrunning) goto escape; // we've harvested all jobs
     }

branches/eam_branches/ipp-20140904/Ohana/src/libdvo/src/dvo_convert.c

@@ -258 +258 @@
     case DVO_FORMAT_##FORMAT: { \
       Measure_##TYPE *tmpMeasure; \
-      myAssert (ABS_COORDS || average, "conversion to internal needs average table"); \
+      myAssert (ABS_COORDS || average, "conversion from internal needs average table"); \
       tmpMeasure = MeasureInternalTo_##TYPE (average, measure, Nmeasure); \
       gfits_table_set_Measure_##TYPE (ftable, tmpMeasure, Nmeasure); \