write a specific error message when the number of elements to gather …

…in PMMG_gather_parmesh exceeds INT_MAX This problem occurs when the size of the first np-1 of the compressed mesh parts exceeds INT_MAX. This commit introduces an error message for this circumstance, which replaces an error on a negative allocation size. It also shows the allocation size in the error message for allocation failur so that this kind of problem is easier to diagnose.
MmgTools · May 18, 2024 · dd37e48 · dd37e48
1 parent 4706e63
commit dd37e48
Show file tree

Hide file tree

Showing 2 changed files with 17 additions and 5 deletions.
diff --git a/src/mergemesh_pmmg.c b/src/mergemesh_pmmg.c
@@ -1098,7 +1098,7 @@ int PMMG_gather_parmesh( PMMG_pParMesh parmesh,
                          int **rcv_next_node_comm,
                          PMMG_pExt_comm **rcv_ext_node_comm ) {
 
-  size_t     pack_size_tot;
+  size_t     pack_size_tot,next_disp;
   int        *rcv_pack_size,ier,ier_glob,k,*displs,ier_pack;
   int        nprocs,root,pack_size;
   char       *rcv_buffer,*buffer,*ptr;
@@ -1145,7 +1145,17 @@ int PMMG_gather_parmesh( PMMG_pParMesh parmesh,
     displs[0] = 0;
     for ( k=1; k<nprocs; ++k ) {
       assert ( displs[k-1] <= INT_MAX - rcv_pack_size[k-1] && "INT_MAX overflow");
-      displs[k] = displs[k-1] + rcv_pack_size[k-1];
+      next_disp = displs[k-1] + rcv_pack_size[k-1];
+      if(next_disp>INT_MAX){
+        /* The displacements argument to MPI_Gatherv() is an array of int
+         * (signed) so the number of elements must be smaller than 2^31.
+         * To get around this we must pack more data in a single element
+         * or use multiple messages.
+         */
+        fprintf(stderr, "  ## Error: too many elements for MPI_Gatherv()\n");
+        MPI_Abort(parmesh->comm, 1);      /* error detected only on root */
+      }
+      displs[k] = next_disp;
     }
     pack_size_tot        = (size_t)(displs[nprocs-1])+(size_t)(rcv_pack_size[nprocs-1]);
     assert ( pack_size_tot < SIZE_MAX && "SIZE_MAX overflow" );

diff --git a/src/parmmg.h b/src/parmmg.h
@@ -289,12 +289,14 @@ static const int PMMG_MVIFCS_NLAYERS = 2;
 
 
 #define ERROR_AT(msg1,msg2)                                          \
-  fprintf( stderr, msg1 msg2 " function: %s, file: %s, line: %d \n", \
-           __func__, __FILE__, __LINE__ )
+  fprintf( stderr, "%s %s function: %s, file: %s, line: %d \n", \
+           msg1, msg2, __func__, __FILE__, __LINE__ )
 
 #define MEM_CHK_AVAIL(mesh,bytes,msg) do {                            \
   if ( (mesh)->memCur + (bytes) > (mesh)->memMax ) {                  \
-    ERROR_AT(msg," Exceeded max memory allowed: ");      \
+    char diag[1024];                                                  \
+    snprintf(diag, 1024, " Allocation of %ld bytes exceeds max %ld: ", bytes, (mesh)->memMax); \
+    ERROR_AT(msg, diag);                                              \
     stat = PMMG_FAILURE;                                              \
   } else if ( (mesh)->memCur + (bytes) < 0  ) {                       \
     ERROR_AT(msg," Tried to free more mem than allocated: " );        \