Just so you know, gsoap is able to be built as an apache module right out of the box, which does what you are looking for (so check out the docs and then explore the generated source a bit), but the apache module will almost certainly be more efficient than a fastcgi connection. If you absolutely cannot use apache, then I guess that's your best bet.
For what it is worth, I found that gsoap will quite happily deal with raw xml arriving in its input buffer without my having to do anything special at all. In fact, it took me awhile to debug what the problem was when I started talking to an external service, since I had my client accidentally configured to leave the http wrapper out of its request and my internal server had been handling it just fine. (just leave the 'http://' off the front of the soap_endpoint url and it won't send the http content). --sam On Mon, Jul 21, 2008 at 7:34 PM, Adam Nielsen <[EMAIL PROTECTED]> wrote: > Hi all, > > I'm in the process of switching from the apparently dead cSOAP project > to gSOAP. > > I've written a small multithreaded FastCGI client to handle the HTTP > side of things (one of the problems with cSOAP was you could crash its > internal HTTP server so easily, so now I'm using a proper HTTP server > and attaching to it via FastCGI) but I'm not quite sure how to attach to > gSOAP *after* the HTTP processing is done. > > Basically I just need to figure out what functions to call to pass in > the raw SOAP XML, without any HTTP headers. The XML is available in a > C++ stream, but I can buffer this into a char array if there's no > alternative. > > At the moment I'm calling soapObj.serve(), but this seems to require the > HTTP headers to be present (it segfaults if they're missing, so I have > to connect with two sets of HTTP headers - one for the webserver and one > for serve().) > > Does anyone know which parts of gSOAP can be used to process messages > normally, but without HTTP headers present? Ideally I'm after something > simple like this: > > soapObj.process(cin, cout); > > Any suggestions? > > Many thanks, > Adam. > >
