Offhand I'd say your compiler might be doing something unexpected with
the math. From your code fragment it looks like the variables are all
supposed to be longs, which ought to work out just fine, but maybe not...

You've got "long xfer = nmemb + size;", where both nmemb and size are
'size_t'. Just picking a system at random here, I've got a Solaris box
that defines size_t:

    typedef unsigned int size_t;

so xfer would be the sum of two unsigned ints - the compiler ought to
cast that to long (I think) as it stores the result, or even before
doing the addition. On the other hand, it might be optimizing to the
point where it doesn't bother storing xfer (local variable, discarded
when the function exits?) but simply substitutes "nmemb + size" wherever
xfer occurs.

If this is the case, when you rearrange that statement you're altering
where the 'long's and 'unsigned int's would be evaluated. Your original
statement was comparing the (supposedly 'long') sum of two unsigned ints
against the sum of two longs. Rearranged, it's comparing the
(supposedly 'long') sum of two unsigned ints plus a long against a long,
and maybe the compiler really does cast everything to the same type at
that point.

I'm not a compiler guru, though, and the above is pure speculation on a
subject that I know almost nothing about. If anyone wants to discuss it
further, let's take it off the list...

Ralph Mitchell

Dhanvi Kapila wrote:

>Hi
>Thanks for the reply. Yeah the "size" in the structure is actually the size
>of the binary data. I already have a custom built webserver that accepts the
>encrypted string.
>
>I finally managed to get solve the problem..though I cannot figure out
>anything wrong in the function logic that I have used below.. The change I
>did in my new code is as shown
>
>
>>> if( xfer >= (in->cbs.size - in->offset) )
>>>
>>>
>changed to
> if( in->offset + xfer >= in->cbs.size )
>
>This change and Curl works like a song .. Any ideas as to what went wrong ?
>
>with regards,
>Dhanvi Kapila
>
>----- Original Message -----
>From: "Ralph Mitchell" <rmitchell_at_eds.com>
>To: <curl-users_at_lists.sourceforge.net>
>Sent: Thursday, July 10, 2003 12:31 PM
>Subject: Re: problems with CURLOPT_READFUNCTION
>
>
>
>
>>This is just a wild guess, you understand... If you're trying to post
>>encrypted data to a web server, you probably need to encode it so that
>>it's more-or-less pure alphanumeric ASCII, otherwise the web server
>>might do something silly with it. Curl doesn't do this for you, so
>>you'll need to find something (or write something) to do it for you.
>>
>>Or it might be a different problem altogether... You could check it out
>>by temporarily passing a simple ASCII string, such as "ABCDEFGH12345678"
>>and see if it makes it though intact.
>>
>>BTW, even if your encrypted data is all alphanumeric, if you're sending
>>a struct that holds a 'size' int you're sending binary...
>>
>>Ralph Mitchell
>>
>>
>>Dhanvi Kapila wrote:
>>
>>
>>
>>>Hi Friends ,
>>>
>>>I am new to the list..
>>>
>>>I am using cURL here at my work place for sending and reading data
>>>
>>>
>between a
>
>
>>>client and server. All the data interaction between is encrypted.
>>>
>>>I am using the CURLOPT_READFUNCTION option of cURL api to send data to
>>>
>>>
>the
>
>
>>>server.
>>>I am provding below a snippet of my code for reference.. I am not sure
>>>
>>>
>where
>
>
>>>I am goin wrong in ... when the server reads the data .. I get a "could
>>>
>>>
>not
>
>
>>>decrypt packet" error
>>>
>>>-- code snippet -----
>>>
>>>typedef struct CBIGSTRING_T
>>>{
>>> CBigString cbs; /* this contains 2 fields.. 1) char *inputString, 2)
>>>long size.. The "inputString" contains the actual encrypted data and
>>>
>>>
>"size"
>
>
>>>is the size of the payload */
>>> long offset;
>>> pool *p; /* Apache resource pool */
>>>}CBigString_t;
>>>
>>>CBigString_t cbst_in ;
>>>
>>>/* filling in the variale with the encrypted content. */
>>>cbst_in.cbs.inputString = (char *)malloc( totalRead +1); /* allocating
>>>storage */
>>>cbst_in.cbs.inputString = (char *)memcpy( cbst_in.cbs.inputString ,
>>>actualStr , totalRead ); /*copying the encrypted data as is from the
>>>tempBuffer given to me by Apache*/
>>>cbst_in.cbs.inputString[totalRead] = '\0'; /* terminating the encrypted
>>>String*/
>>>cbst_in.cbs.size = totalRead;
>>>cbst_in.offset = 0;
>>>cbst_in.p = p;
>>>
>>>curl_easy_setopt( realmHandle , CURLOPT_READFUNCTION ,
>>>Curl_SendDataToServer ); /* function to send data to the server*/
>>>curl_easy_setopt( realmHandle , CURLOPT_READDATA , &cbst_in );
>>>
>>>
>/*
>
>
>>>variable contains the encrypted packet */
>>>
>>>static int Curl_SendDataToServer(void *buffer, size_t size, size_t nmemb,
>>>CBigString_t *in)
>>>{
>>> LogTrace( in->p , " ----------> Entering Curl_SendDataToServer
>>><----------");
>>> if (in == NULL) {
>>> return -1;
>>> }
>>>
>>> long xfer = nmemb * size ;
>>> /* Check if the size presented by CURL is greated than that actually
>>>needs to be copied */
>>> LogDebug( in->p , " [ OFFSET, SIZE ] => [ %d , %d ] ", in->offset ,
>>>in->cbs.size);
>>>
>>> if( xfer >= (in->cbs.size - in->offset) )
>>> {
>>> xfer = (in->cbs.size - in->offset) ;
>>> }
>>> if( in->offset < in->cbs.size )
>>> {
>>> buffer = memcpy( (char *)buff, in->cbs.inputString + in->offset,
>>>xfer );
>>> in->offset += xfer; /* reflect the updated offset value.. which
>>>
>>>
>says
>
>
>>>stores how much has been read from the encrypted payload. */
>>> }
>>> else
>>> {
>>> xfer = -1;
>>> }
>>> LogDebug( in->p , " BUFFER SIZE : %d, %d , |%s| ", size*nmemb, xfer,
>>>buffer);
>>> LogTrace( in->p , " ----------> Leaving Curl_SendDataToServer
>>><----------");
>>> return xfer;
>>>}
>>>---- end of code -----
>>>
>>>when I do a tcpdump of the arrived packet and the dispatched packet.. the
>>>packet data dont match.. hence the error is when I am sending the data
>>>
>>>
>using
>
>
>>>this piece of code.. please help me.. I am at a loss to find out the
>>>error....
>>>
>>>Awaiting your replies.. Thank you for your help and time.
>>>
>>>-Dhanvi Kapila
>>>
>>>
>>>-------------------------------------------------------
>>>This SF.Net email sponsored by: Parasoft
>>>Error proof Web apps, automate testing & more.
>>>Download & eval WebKing and get a free book.
>>>www.parasoft.com/bulletproofapps
>>>
>>>
>>>
>>>
>>
>>
>>-------------------------------------------------------
>>This SF.Net email sponsored by: Parasoft
>>Error proof Web apps, automate testing & more.
>>Download & eval WebKing and get a free book.
>>www.parasoft.com/bulletproofapps
>>
>>
>>
>
>
>-------------------------------------------------------
>This SF.Net email sponsored by: Parasoft
>Error proof Web apps, automate testing & more.
>Download & eval WebKing and get a free book.
>www.parasoft.com/bulletproofapps1
>
>

-------------------------------------------------------
This SF.Net email sponsored by: Parasoft
Error proof Web apps, automate testing & more.
Download & eval WebKing and get a free book.
www.parasoft.com/bulletproofapps1
Received on 2003-07-15