Re: [gmime-devel] Problem parsing HTTP POST multipart/form-data

[Jeffrey Stedfast <fejj gnome org>]

Hi Ron,

On 4/15/2016 4:48 AM, Ron wrote:
> Hi,
>
> I'm having huge problems parsing multipart/form-data. I'm using 2.4.21
> (because CentOS 6).
>
> This is my POST command:
>
> curl -v -F key1=value1 -F "key2=value2" -F "file= test txt"
> http://localhost:5657/v1/test
>
> This is the resulting HTTP request:

[snip]

> Now, how do I parse this? On the server side I'm using libevent. My
> approach was:
>
>    stream = g_mime_stream_mem_new();
>
>    // not shown, but write all headers to the stream using
> g_mime_stream_write() each followed by a CRLF
>    // add empty CRLF, and add the body using g_mime_stream_write()

Don't forget to call g_mime_stream_reset (stream);

Otherwise the parser will attempt to parse the stream starting at the 
very end of the content (i.e. the parser will have nothing to parse).

>    parser = g_mime_parser_new_with_stream (stream);
>
>    // parse the block from the stream.  vdoes not exist?
>    multipart = g_mime_multipart_new();
>    while ((object = g_mime_parser_construct_part(parser)) != NULL) {
>      g_mime_multipart_add(multipart, object);
>    }
>    // but this loops only once??

Yes, it will only loop once because it will return a GMimeMultipart 
since that's what the content is.

The HTTP headers contain the following Content-Type header:

Content-Type: multipart/form-data;
 boundary=----------------------------2912a7dc5b77


and so GMimeParser will parse that header and recognize that it is 
parsing a multipart and so create a new GMimeMultipart object and then 
parse the boundary string and use that to figure out where each of the 
child parts begin and end, appending them into the multipart/form-data 
object that it instantiated once it parsed the Content-Type header.

When it finishes parsing all of the children of the multipart/form-data 
part, it will return the multipart/form-data object.

Hope that helps,

Jeff