[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[netCDF #PKT-462504]: Re: [netcdfgroup] netcdf-4 open/close memory leak

> 
> Hi, Ed,
> 
> I installed valgring and ran the tests (on OS X) on a 32-bit compile
> with -g -O0 (for both hdf5 and netcdf), and the netcdf "make check"
> didn't find any problems.  Nor does valgrind find any errors with Jeff
> Whitaker's simple program (attached, along with his test netcdf
> files).  But if the iteration count is large enough to keep the
> program running for a while, top clearly shows the memory usage
> climbing about linearly with time.  The memory usage goes up for both
> 32 and 64 bit compilations, for what it's worth.
> 
> I don't have much experience with hdf5, but I did modify the hdf5
> example program "h5_read" to run in a big loop (with no printing and
> everything stripped out execept H5Fopen/H5Fclose), and it doesn't show
> any obvious leakage in top.
> 
> -- Ted
> 
> 

Howdy Ted!

I have been looking into memory issues. I have changed libsrc4/tst_files2.c to 
get some actual numbers, but I wonder if you might want to try calling this 
function in some of your code to see what results you get:
/* This function uses the ps command to find the amount of memory in            
                                                                                
                 
use by the process. From the ps man page:                                       
                                                                                
                 
                                                                                
                                                                                
                 
size SZ approximate amount of swap space that would be required if              
                                                                                
                 
        the process were to dirty all writable pages and then be                
                                                                                
                 
        swapped out. This number is very rough!                                 
                                                                                
                 
*/
void
get_mem_used(int *mem_used)
{
   char cmd[NC_MAX_NAME + 1];
   char blob[MAX_LEN + 1];
   FILE *fp;
   int num_char;

   sprintf(cmd, "ps -o size= %d > %s", getpid(), TMP_FILE_NAME);
   system(cmd);
   if (!(fp = fopen(TMP_FILE_NAME, "r"))) ERR;
   num_char = fread(blob, MAX_LEN, 1, fp);
   sscanf(blob, "%d", mem_used);
   fclose(fp);
   unlink(TMP_FILE_NAME);
}



Ticket Details
===================
Ticket ID: PKT-462504
Department: Support netCDF
Priority: Critical
Status: Closed