FreeTDS Development Group

[miscmail <miscmail AT fingerpost.co.uk>]

Hi all

I am trying to upgrade from freetds 0.62 to 0.82 and have a small  
problem with dbwritetext

In my program, if the data length < system_max then the program does a :

 - dbwritetext with a data buffer set to NULL and the length set to  
the actual size.
 - then a loop with dbmoretext to upload the chunks

But in 0.82 it just hangs ... with an error .....

In dblib/unittest/t0013.c there is some code which is similar to mine  
BUT it is wasted with a ''ifndef"

When I tried a file around 400k - it worked fine even though the 'set  
textsize' is  64512

So then I set the flag (and added missing variables numtowrite and  
numwritten) :
        #define DBWRITE_OK_FOR_OVER_4K

.. recompiled and got the same hang as my code does

The actual code in t0013 is :

        if (dbwritetext(dbproc, objname, textPtr, DBTXPLEN,  
timeStamp, FALSE, isiz, NULL) != SUCCEED)
                return 15;
        dbsqlok(dbproc);
        dbresults(dbproc);

        numtowrite = 0;
        /* Send the update value in chunks. */
        for (numwritten = 0; numwritten < isiz; numwritten +=  
numtowrite) {
                numtowrite = (isiz - numwritten);
                if (numtowrite > BLOB_BLOCK_SIZE)
                        numtowrite = BLOB_BLOCK_SIZE;
                dbmoretext(dbproc, (DBINT) numtowrite, blob +  
numwritten);
        }
        dbsqlok(dbproc);
        while (dbresults(dbproc) != NO_MORE_RESULTS);

I checked the src/dblib/dblib.c an found there is a check-exit-if-null  
on the 'text' buffer parameter
- around line 6261 in dblib.c 0.82 there is a
        CHECK_NULP(text, "dbwritetext", 8, FAIL);
        which looks like it is doing the damage.

which means a call with a NULL text buffer never gets to do anything !

There is a snippet of code a few lines further down which catches  
NULLs (below) so I wonder if that CHECK is superfluous :

        if (!text) {
                dbproc->text_size = size;
                dbproc->text_sent = 0;
                return SUCCEED;
        }

Is there something new in how dbwritetext works - or should I just  
comment out the offending CHECK_NULP line  (BTW it aint in freetds  
0.62)?

Waddatink

chris