test: add impls/mpich/ulfm/get_failed.c

This is adapted from ft/failure_ack.c. The current ft tests are disabled
by wholesale. Add to impls to get it tested. MPIX functions need be in
impls folder anyway.

test/mpi/configure.ac

@@ -1692,5 +1692,6 @@ AC_OUTPUT(maint/testmerge \
           impls/mpich/threads/pt2pt/Makefile \
           impls/mpich/cuda/Makefile \
           impls/mpich/misc/Makefile \
+          impls/mpich/ulfm/Makefile \
           )
 

test/mpi/impls/mpich/Makefile.am

@@ -7,5 +7,5 @@ include $(top_srcdir)/Makefile_single.mtest
 
 EXTRA_DIST = testlist.in
 
-static_subdirs = mpi_t comm misc
+static_subdirs = mpi_t comm misc ulfm
 SUBDIRS      = $(static_subdirs) $(threadsdir) $(cudadir)

test/mpi/impls/mpich/ulfm/Makfile.am

@@ -0,0 +1,11 @@
+##
+## Copyright (C) by Argonne National Laboratory
+##     See COPYRIGHT in top-level directory
+##
+
+include $(top_srcdir)/Makefile_single.mtest
+
+EXTRA_DIST = testlist
+
+noinst_PROGRAMS = \
+    get_failed

test/mpi/impls/mpich/ulfm/get_failed.c

@@ -0,0 +1,115 @@
+/*
+ * Copyright (C) by Argonne National Laboratory
+ *     See COPYRIGHT in top-level directory
+ */
+
+#include <mpi.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+/*
+ * This test makes sure that after a failure, the correct group of failed
+ * processes is returned from MPIX_Comm_get_failed.
+ */
+int main(int argc, char **argv)
+{
+    int rank, size, err, result;
+    char buf[10] = " No errors";
+    MPI_Group failed_grp, world_grp;
+    int world_ranks[] = { 0, 1, 2 };
+    int failed_ranks[3];
+
+    MPI_Init(&argc, &argv);
+    MPI_Comm_rank(MPI_COMM_WORLD, &rank);
+    MPI_Comm_size(MPI_COMM_WORLD, &size);
+    if (size < 3) {
+        fprintf(stderr, "Must run with at least 3 processes\n");
+        MPI_Abort(MPI_COMM_WORLD, 1);
+    }
+
+    MPI_Comm_group(MPI_COMM_WORLD, &world_grp);
+    MPI_Comm_set_errhandler(MPI_COMM_WORLD, MPI_ERRORS_RETURN);
+
+    if (rank == 1) {
+        exit(EXIT_FAILURE);
+    }
+
+    if (rank == 0) {
+        err = MPI_Recv(buf, 10, MPI_CHAR, 1, 0, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
+        if (MPI_SUCCESS == err) {
+            fprintf(stderr, "Expected a failure for receive from rank 1\n");
+            MPI_Abort(MPI_COMM_WORLD, 1);
+        }
+
+        err = MPIX_Comm_get_failed(MPI_COMM_WORLD, &failed_grp);
+        if (MPI_SUCCESS != err) {
+            int ec;
+            char error[MPI_MAX_ERROR_STRING];
+            MPI_Error_class(err, &ec);
+            MPI_Error_string(err, error, &size);
+            fprintf(stderr, "MPIX_Comm_get_failed returned an error: %d\n%s", ec, error);
+            MPI_Abort(MPI_COMM_WORLD, 1);
+        }
+
+        MPI_Group one_grp;
+        int one[] = { 1 };
+        MPI_Group_incl(world_grp, 1, one, &one_grp);
+        MPI_Group_compare(one_grp, failed_grp, &result);
+        if (MPI_IDENT != result) {
+            fprintf(stderr, "First failed group contains incorrect processes\n");
+            MPI_Group_size(failed_grp, &size);
+            MPI_Group_translate_ranks(failed_grp, size, world_ranks, world_grp, failed_ranks);
+            for (int i = 0; i < size; i++)
+                fprintf(stderr, "DEAD: %d\n", failed_ranks[i]);
+            MPI_Abort(MPI_COMM_WORLD, 1);
+        }
+        MPI_Group_free(&failed_grp);
+        MPI_Group_free(&one_grp);
+
+        err = MPI_Recv(buf, 10, MPI_CHAR, 2, 0, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
+        if (MPI_SUCCESS != err) {
+            fprintf(stderr, "First receive failed\n");
+            MPI_Abort(MPI_COMM_WORLD, 1);
+        }
+        err = MPI_Recv(buf, 10, MPI_CHAR, 2, 0, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
+        if (MPI_SUCCESS == err) {
+            fprintf(stderr, "Expected a failure for receive from rank 2\n");
+            MPI_Abort(MPI_COMM_WORLD, 1);
+        }
+
+        err = MPIX_Comm_get_failed(MPI_COMM_WORLD, &failed_grp);
+        if (MPI_SUCCESS != err) {
+            int ec;
+            char error[MPI_MAX_ERROR_STRING];
+            MPI_Error_class(err, &ec);
+            MPI_Error_class(err, &ec);
+            MPI_Error_string(err, error, &size);
+            fprintf(stderr, "MPIX_Comm_get_failed returned an error: %d\n%s", ec, error);
+            MPI_Abort(MPI_COMM_WORLD, 1);
+        }
+
+        MPI_Group two_grp;
+        int two[] = { 1, 2 };
+        MPI_Group_incl(world_grp, 2, two, &two_grp);
+        MPI_Group_compare(two_grp, failed_grp, &result);
+        if (MPI_IDENT != result) {
+            fprintf(stderr, "Second failed group contains incorrect processes\n");
+            MPI_Group_size(failed_grp, &size);
+            MPI_Group_translate_ranks(failed_grp, size, world_ranks, world_grp, failed_ranks);
+            for (int i = 0; i < size; i++)
+                fprintf(stderr, "DEAD: %d\n", failed_ranks[i]);
+            MPI_Abort(MPI_COMM_WORLD, 1);
+        }
+        MPI_Group_free(&failed_grp);
+        MPI_Group_free(&two_grp);
+
+        fprintf(stdout, " No errors\n");
+    } else if (rank == 2) {
+        MPI_Ssend(buf, 10, MPI_CHAR, 0, 0, MPI_COMM_WORLD);
+
+        exit(EXIT_FAILURE);
+    }
+
+    MPI_Group_free(&world_grp);
+    MPI_Finalize();
+}

test/mpi/impls/mpich/ulfm/testlist

@@ -0,0 +1 @@
+get_failed 3 mpiexecarg=-disable-auto-cleanup resultTest=TestStatusNoErrors timeLimit=10