visit_broadcast_string_callback detects a negative string lenght for a parallel simulation.

[MuzafarWani]

I am tapping into a simulation data by fetching data from NetCDF functions in the simulation. For one rank the below loop works perfectly fine

_void controlloop(simulation_data *sim, int timestep, int *cts)
{
int blocking, visitstate = 0;

// sim->cycle = *cts;
// printf("DS: Timestep is %d\n",timestep);
if (sim->par_rank == 0)
{
fprintf(stderr, "command> ");
fflush(stderr);
}

do
{
      if (sim->runMode == SIM_STOPPED) 
          blocking = 0;          
      else 
          blocking = 1;


    /* Get input from VisIt or timeout so the simulation can run. */
    
     visitstate = VisItDetectInput(blocking, fileno(stdin));
   //  printf("Visit: visitstate = %d\n",visitstate);
     
    /* Do different things depending on the output from VisItDetectInput. */
    if(visitstate >= -5 && visitstate <= -1)
    {
        fprintf(stderr, "Can't recover from error!\n");
        //err = 1;
    }
    else if(visitstate == 0)
    {
        /* There was no input from VisIt, return control to sim. */
        simulate_one_timestep(sim,cts);
    }
    else if(visitstate == 1)
    {    printf("Visit:Inside\n");
        /* VisIt is trying to connect to sim. */
        if(VisItAttemptToCompleteConnection() == VISIT_OKAY)
        {   
            fprintf(stderr, "VisIt connected\n");
            printf("Visit:Inside1\n");
            VisItSetCommandCallback(ControlCommandCallback, (void*)sim);  
            VisItSetGetMetaData(SimGetMetaData, (void*)sim);
            VisItSetGetMesh(SimGetMesh, (void*)sim);
            VisItSetGetVariable(SimGetVariable, (void*)sim);
            printf("Visit : visflag =%d\n",sim->visflag);  
            
        }
        else
            fprintf(stderr, "VisIt did not connect\n");
    }
    else if(visitstate == 2)
    {
        
        if(VisItProcessEngineCommand() == VISIT_ERROR)
        {
            /* Disconnect on an error or closed connection. */
            VisItDisconnect();
            /* Start running again if VisIt closes. */
            sim->runMode = SIM_RUNNING;
        }
    }
    else if(visitstate == 3)
    {
        ProcessConsoleCommand(sim);
        fprintf(stderr, "command> ");
        fflush(stderr);
    }
    
    if(sim->visflag == 3)
    *cts = *cts + 1;
 // printf("Visit : visflag =%d\n",sim->visflag);  
} while(sim->visflag < 3);

}_

But once I add the code for more than 1 rank. It throws errors from the below function

static int visit_broadcast_string_callback(char *str, int len, int sender)
{
return MPI_Bcast(str, len, MPI_CHAR, sender, MPI_COMM_WORLD);
}

it detects a data len either as a negative number or a very large number which has no relation to the simulation.
I have changed the loop to below for more than 1 rank

_void controlloop(simulation_data *sim, int timestep, int *cts)
{
int blocking, visitstate = 0;

// sim->cycle = *cts;
// printf("DS: Timestep is %d\n",timestep);
if (sim->par_rank == 0)
{
fprintf(stderr, "command> ");
fflush(stderr);
}

do
{
      if (sim->runMode == SIM_STOPPED) 
          blocking = 0;          
      else 
          blocking = 1;


    /* Get input from VisIt or timeout so the simulation can run. */
    
     visitstate = VisItDetectInput(blocking, fileno(stdin));
   
     
    /* Do different things depending on the output from VisItDetectInput. */
    if(visitstate >= -5 && visitstate <= -1)
    {
        fprintf(stderr, "Can't recover from error!\n");
        //err = 1;
    }
    else if(visitstate == 0)
    {
        /* There was no input from VisIt, return control to sim. */
        simulate_one_timestep(sim,cts);
    }
    else if(visitstate == 1)
    {    printf("Visit:Entry 1\n");
        /* VisIt is trying to connect to sim. */
        if(VisItAttemptToCompleteConnection() == VISIT_OKAY)
        {   
            fprintf(stderr, "VisIt connected\n");
            printf("Visit:Entry 2\n");
            VisItSetCommandCallback(ControlCommandCallback, (void*)sim);  
            VisItSetSlaveProcessCallback2(SlaveProcessCallback, (void*)sim);
            
            VisItSetGetMetaData(SimGetMetaData, (void*)sim);
            VisItSetGetMesh(SimGetMesh, (void*)sim);
            VisItSetGetVariable(SimGetVariable, (void*)sim);
            VisItSetGetDomainList(SimGetDomainList, (void*)sim);
          
            printf("Visit : visflag =%d\n",sim->visflag);  
            
        }
        else
            fprintf(stderr, "VisIt did not connect\n");
    }
    else if(visitstate == 2)
    {
        
        if(VisItProcessEngineCommand() == VISIT_ERROR)
        {
            /* Disconnect on an error or closed connection. */
            VisItDisconnect();
            /* Start running again if VisIt closes. */
            sim->runMode = SIM_RUNNING;
        }
    }
    else if(visitstate == 3)
    {
        ProcessConsoleCommand(sim);
        fprintf(stderr, "command> ");
        fflush(stderr);
    }
    
    if(sim->visflag == 3)
    *cts = *cts + 1;
 // printf("Visit : visflag =%d\n",sim->visflag);  
} while(sim->visflag < 3);

}_

The above loop never enters the condition

if(VisItAttemptToCompleteConnection() == VISIT_OKAY)

but it does pass the condition check
else if(visitstate == 1)

It is able to detect the incoming connection but doesn't complete the same.

Can you let me know how to resolve it

[MuzafarWani]

@markcmiller86 @biagas @brugger1

[biagas]

static int visit_broadcast_string_callback(char *str, int len, int sender)
{
return MPI_Bcast(str, len, MPI_CHAR, sender, MPI_COMM_WORLD);
}

Is this your definition of the function? Else where are you getting this definition?
You also haven't shown where you call VisItSetBroadcastStringCallback, and where/how you've initialized MPI and set up the sim's par_com. It's a bit hard to help when I cannot see all the code.

Have you looked at any of the simulation examples for help in setting up your control loop?
The domaindbounds example might be a good one to look at.

I see a difference in your setting of blocking from what I see in the sim examples.
The sim examples I've looked at all set blocking in this way:
blocking = (sim->runMode == SIM_RUNNING) ? 0 : 1;
(0 when the sim is running, otherwise 1).
Your logic sets blocking to 0 when the sim is stopped, otherwise 1.

[biagas]

@MuzafarWani Comments you added on January 5th seem to have been lost. I received the email, but they never showed up on github.
Could you repost them in a new discussion since it is a different problem than this one. Also, a discussion (not bug ticket) is the better place for it. Thank you.