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

[netCDF #DMJ-216367]: Reading a dynamic 2d array.

Hi Anton,

The problem is that the address of the memory you pass to the
nc_get_vars_text() function with the &flags[0][0] argument is not a contiguous
block of memory of size tlen*flen, as required by the netCDF interface.
Instead, &flags[0][0] is the address of the first row of the flags array,
allocated separately from the other rows inside the loop
 for( int i = 0; i < tlen; i++)
 flags[i] = new char[flen];
so it is really only a block of length flen.  The next row is allocated with a
different new statement, so could be anywhere in memory, and is not likely to
be contiguous with the first row.  You allocate a toatal of tlen*flen chars,
but they are in tlen separate blocks of length flen each, and they are just
located in the heap somewhere, not in a single block.  The called function,
nc_get_vars_text(), has no way of knowing how the memory buffer is organized in
little separated chunks all over the heap, it can only assume there is a single
block of memory it can read values into.

A fix for this is to use the C++ "placement new operator", which constructs an
object using a pre-allocated buffer.  For example, I think this would work:

//create matrix of correct size.
buf = new char[tlen*flen];  // a contiguous block of memory
flags = new char*[tlen];
for( int i = 0; i < tlen; i++)
flags[i] = new (&flags[i*flen]) char[flen]; // use placement new operator
  ...
// the following line should no longer produce a seg fault:
status = nc_get_vars_text( ncfileid, flagid, start, count, NULL, &flags[0][0]
);

--Russ


Russ Rew                                         UCAR Unidata Program
address@hidden                     http://www.unidata.ucar.edu



Ticket Details
===================
Ticket ID: DMJ-216367
Department: Support netCDF
Priority: High
Status: Closed