Maxim had once mentioned about checking out memcache module for an
example of non-buffered protocol handling, and i modeled my code
around it. Is the behavior below because of buffered protocol (http)?
Thanks,
Ashish
On Wed, Apr 25, 2012 at 5:22 PM, Ashish S <
[email protected]> wrote:
> Hi there,
>
> In my module which uses upstream to talk to a backend, i do the
> following in the response processing method (u->process_header)
>
> {
> ngx_http_upstream_t * u = r->upstream;
> size_t myResponseLength = (u->buffer.last - u->buffer.pos);
> size_t origUpstreamResponseLength = (u->buffer.last - u->buffer.start);
> std::string myResponseString ((u_char *) u->buffer.pos, myResponseLength);
>
> ngx_log_error(NGX_LOG_ERR, r->connection->log, 0, "u->length is %d
> and Original Upstream response length is %d and myResponseLength is
> %d", u->length, origUpstreamResponseLength, myResponseLength);
>
> ....
> }
>
> My actual response length from upstream is ~16kb, while i get
> u->length printed as 0, Original upstream response length and
> myResponseLength printed as 1448 bytes. I also see in nginx debug
> logs, "Transfer-Encoding: chunked". Is the difference in length
> printed, because of this encoding? Is there a way to disable chunked
> transfer, so that i can get the entire response in one go (into my
> string), as shown in code? If not, what is the right method to read
> the entire response from upstream buffer into a temporary string (so
> that i can do further parsing)?
>
> Thanks,
> Ashish
_______________________________________________
nginx mailing list
nginx@nginx.org
http://mailman.nginx.org/mailman/listinfo/nginx