-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On 04/26/2012 05:50 AM, Martin Dluhos wrote:
> On 04/25/2012 09:22 AM, Christian Grothoff wrote:
>> The first call to the "test_send_response_helper" is made *before*
>> the upload data is available (after we just got the headers) so
>> that you can choose not to send 100 CONTINUE for http/1.1. You
>> need to change your function to return an 'int' and return
>> 'MHD_YES' to get the next call which will contain upload data.
>> Furthermore, 'upload_data' is NOT a 0-terminated string so you
>> *must* limit your processing manually to the first
>> '*upload_data_size' bytes in that buffer. Finally, you then need
>> to reduce '*upload_data_size' (i.e. to zero) to indicate how many
>> bytes of 'upload_data' you did process (so you don't get them
>> again). Finally, if you are eventually called a *second* time with
>> '0 == *upload_data_size', you can be sure that the upload is
>> complete.
>
> Thank you for your help, Christian. I forgot that the function behaves
> in this way.
>
> I actually have a follow up question. Inside test_send_response_helper
> I make a call to a function called send_response, which uses
> microhttpd library functions to create and send an MHD_Response. Here
> is the code:
>
> int
> test_send_response_helper (void *cls, struct MHD_Connection *connection,
> const char *url, const char *method,
> const char *version, const char *upload_data,
> size_t *upload_data_size, void **con_cls)
> {
> int ret_val;
>
> /* The first time through only headers are processed. Only the second
> * time this function is called can we access upload_data.
> */
> if (*con_cls == NULL)
> {
> struct connection_info_struct *con_info;
> con_info = malloc (sizeof (struct connection_info_struct));
>
> /* Process POST and GET request separately. Signal an error
> * on any other method.
> */
> if (strcmp (method, "POST") == 0)
> con_info->connection_type = POST;
> else if (strcmp (method, "GET") == 0)
> con_info->connection_type = GET;
>
> *con_cls = (void *) con_info;
> return MHD_YES;
> } // if
>
> /* This gets executed the second time the function is called. Only
> * now can we access upload_data.
> */
> printf ("SECOND: URL: %s DATA: %s\n", url, upload_data);
>
> /* Response strings for the server to return. */
> const char *busy_page =
> "The server is too busy to handle the verification request.";
>
> /* The server is too busy to respond. */
> ret_val = send_response(connection, busy_page,
> MHD_HTTP_SERVICE_UNAVAILABLE);
>
> return MHD_YES;
> }
>
> int
> send_response (struct MHD_Connection *connection, const char
> *response_data, int status_code)
> {
> int return_value;
> struct MHD_Response *response;
>
> response = MHD_create_response_from_data (strlen(response_data),
> (void*) response_data, MHD_NO, MHD_NO);
>
> if (response == NULL)
> {
> return MHD_NO;
> }
>
> return_value = MHD_queue_response (connection, status_code, response);
> MHD_destroy_response (response);
>
> return return_value;
> }
>
> For some reason, MHD_queue_response returns MHD_NO indicating an error
> instead of MHD_YES. Is there a way for me to find out what kind of
> error I am getting here? Can you see what I am doing wrong here?
Yes, there are only two times where you're allowed to queue a response
- ---either the FIRST time you're called (for queueing errors like service
unavailable). Those should be done *before* you receive upload data.
Alternatively, you must wait until you have received *all* upload data
and then queue your response once "*upload_data_size" is *again* zero.
You don't wait for the upload to finish, so MHD tells you that you're
using the API the wrong way. The only error you get from
MHD_queue_response is pretty much "API misuse" (called too early, called
twice).
Happy hacking!
Christian
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
iEYEARECAAYFAk+Y4NYACgkQv2Bwi0hCbH5jrwCfeVcjMP01kJZanRFIlvcnHy1B
8V8AoJ2oAD1bWGoI9IeHXM7jBlvSpjmA
=LYY2
-----END PGP SIGNATURE-----
