Work around a testphdf5 failure on Cray MPICH machines (HDFGroup#3361)

release_docs/RELEASE.txt

@@ -260,6 +260,17 @@ Bug Fixes since HDF5-1.14.1 release
 
     Testing
     -------
+    - Fixed a testing failure in testphdf5 on Cray machines
+
+      On some Cray machines, what appears to be a bug in Cray MPICH was causing
+      calls to H5Fis_accessible to create a 0-byte file with strange Unix
+      permissions. This was causing an H5Fdelete file deletion test in the
+      testphdf5 program to fail due to a just-deleted HDF5 file appearing to
+      still be accessible on the file system. The issue in Cray MPICH has been
+      worked around for the time being by resetting the MPI_Info object on the
+      File Access Property List used to MPI_INFO_NULL before passing it to the
+      H5Fis_accessible call.
+
     - A bug was fixed in the HDF5 API test random datatype generation code
 
       A bug in the random datatype generation code could cause test failures

testpar/t_file.c

@@ -986,20 +986,23 @@ test_delete(void)
     ret = H5Fdelete(filename, fapl_id);
     VRFY((SUCCEED == ret), "H5Fdelete");
 
+    /*
+     * Work around a Cray MPICH bug that causes
+     * H5Fis_accessible to re-create the just-deleted
+     * file as a 0-byte file with strange Unix
+     * permissions, causing the routine to return
+     * false here instead of FAIL.
+     */
+    H5Pset_fapl_mpio(fapl_id, comm, info);
+
     /* Verify that the file is NO LONGER an HDF5 file */
     /* This should fail since there is no file */
     H5E_BEGIN_TRY
     {
         is_accessible = H5Fis_accessible(filename, fapl_id);
     }
     H5E_END_TRY
-
-    if (FALSE == is_accessible) {
-        VRFY((FALSE == is_accessible), "H5Fis_accessible returned FALSE");
-    }
-    if (FAIL == is_accessible) {
-        VRFY((FAIL == is_accessible), "H5Fis_accessible failed");
-    }
+    VRFY((FAIL == is_accessible), "H5Fis_accessible failed as expected");
 
     /* Release file-access plist */
     ret = H5Pclose(fapl_id);