> I had a chance to look into the code to understand what
> raw_post does and why didn't work (as KL pointed out).
> The reason for raw_post not returning data (I don't have
> tcl errors) is that the command actually is implemented
> only for uurlencode form data. Multipart/form-data parsing
> simply doesn't set a the pointer to a buffer with the
> whole data posted by the client. Actually such a buffer
> doesn't even exist for multipart messages: multipart data
> are read in chunks from the client and parsed as they are
> handed in by the apache framework. Being raw_post chiefly
> a command for debugging (as far as I can understand)
> I imagine there was no necessity to give a full
> implementation and documentation.
>
> with the last commit of TclWebapache.c Karl changed the
> following test in this way
>
> int TclWeb_UploadSave(char *varname, Tcl_Obj *filename, TclWebRequest *req)
> {
> apr_status_t status;
> @@ -569,25 +577,27 @@
> rsc = RIVET_SERVER_CONF( req->req->server->module_config );
> /* This sucks - we should use the hook, but I want to
> get everything fixed and working first */
> - if (rsc->upload_files_to_var)
> + if (1 || rsc->upload_files_to_var)
> {
> - /* char *bytes = NULL; */
>
> ....
>
> disabling the test on upload_files_to_var altogether. I
> suppose this was for test and was meant to be restored
> in its pristine form.
Yeah, 1 || looks like an oops.
What would really be ideal is to beat the test harness back into shape
so that these things are easy to write tests for, and thus ensure that
other stuff doesn't get broken when changes get committed...
--
David N. Welton
http://www.welton.it/davidw/
http://www.dedasys.com/
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
