On Fri, 14 Mar 2008, Dan Petitt wrote:

Please don't top-post.

> I don't know what else to say really, I have got reproducible code that I
> sent you so I cant really do much else.

There are a lot more stuff you can do. For example firing up a debugger, set a
few break-points in the lib and trying to figure out why you get what you get.

And my example should be very easy to get to build and try on windows too.

It isn't like I possess some magic key to the kingdom. Everything is public
and open, available for everyone to join in!

> I havent tested it yet, but the only thought I had was that if you posted
> information that was less than the buffer size then the
> progressupdatecounter doesn't get set and thus it stays at zero on any
> progress updates.

Well don't you see? This is now totally new information that you didn't
mention previously. Yes, my example shows a clear "step effect" in that it
only increases the process counter for uploads with 16K steps, which is the
internal buffer size. Of course in my test I "fake" a slow upload by limiting
the upload speed using the libcurl option for it so I'm not 100% sure what
happens when the actual network speed is lower. But then the topic of this
thread is "for large posts" so I would assume that means sizes greater than
16K.

I'd say it is likely that there are still improvements left to be done in that
area in the code. Decreasing the internal buffer size with CURLOPT_BUFFERSIZE
is also a kind of work-around for it too.

-- 
  Commercial curl and libcurl Technical Support: http://haxx.se/curl.html