Re: Memory Leaks - Can you help me find them in ths snippet

From:

"Daniel T." <daniel_t@earthlink.net>

Newsgroups:

comp.lang.c++

Date:

Tue, 29 Jan 2008 13:54:49 -0500

Message-ID:

<daniel_t-469733.13544829012008@earthlink.vsrv-sjc.supernews.net>

In article <13pukm54dvccv84@corp.supernews.com>,
"Alf P. Steinbach" <alfps@start.no> wrote:

* Daniel T.:

"nmehring@gmail.com" <nmehring@gmail.com> wrote:

Thank you, I will correct that additional leak if i>0

The reason I am using these datatypes is because I am utilizing this
3rd party library method:

extern LONG SDEAPI SE_stream_update_row (SE_STREAM stream,
                                          const ACHAR *table,
                                          LONG *sde_row_id,
                                          SHORT num_columns,
                                          const ACHAR **columns);

So I figured I had to use the char data type.

Sorry, I thought you were using each column individually. i.e., I
thought the API was asking for char* and was being called multiple
times, not char** and called once.

However, I still recommend you use vectors rather than allocating the
memory yourself and hoping for the best.

Something like this would do nicely:

   vector< vector< char > >
                        block( lColumnCount, SE_QUALIFIED_COLUMN_LEN );

   for ( int i = 0; i < block.size(); ++i )
   {
      strcpy( &block[i].front(), CStringColumnName[i] );
   }

   vector< char* > columns( block.size() );
   for ( int i = 0; i != block.size(); ++i )
   {
      columns[i] = &block[i].front();
   }
   SE_stream_update_row( /* other params */, &columns[0] );

This is a better idea than the code I posted: it uses a little more
memory but is safer (wrt. maintainance) and shorter and just more clear.

However, the 'block' constructor arguments need to be fixed, and the
whole thing needs to be put in a try-catch in order to conform to the
OP's boolean return.

Accepted about the try catch block, but what is wrong with the
constructor arguments, did I reverse them or something?

I'd just default-initialize the inner vectors, and .resize() them in the
string copy loop.

Another idea would be to allocate the block as a single vector<char>
rather than a vector of vectors, then partition the block into the
columns vector, like so:

vector< char > block( lColumnCount * SE_QUALIFIED_COLUMN_LEN );
vector< char* > columns( lColumnCount );
   for ( int i = 0, j = 0; i < lColumnCount;
                                    ++i, j += SE_QUALIFIED_COLUMN_LEN ) {
      columns[i] = &block[j];
   }
   // and so on.